fix(react): Add `POP` guard for long-running `pageload` spans by onurtemizkan · Pull Request #17867 · getsentry/sentry-javascript

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan · 2025-10-06T11:09:59Z

This resolves the issue that occurs when an extra navigation transaction is created after a prematurely ended pageload transaction in React Router lazy routes.

This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening).

This causes the pageload transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a POP event emitted, which we subscribe to create a navigation transaction. This ends the ongoing pageload transaction before its name is updated with a resolved parameterized route path, and starts a navigation transaction, which contains the remaining spans that were supposed to be a part of the pageload transaction.

This fix makes sure the initial POP events are not necessarily treated as navigation pointers, which should fix both:

Duplicate / extra navigation transactions having a part of pageload spans.
Remaining wildcards in the pageload transaction names

chargome

Nice find! Any way we can test this in e2e?

github-actions · 2025-10-08T20:05:19Z

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	9,373	-	8,903	+5%
GET With Sentry	1,378	15%	1,367	+1%
GET With Sentry (error only)	6,233	66%	6,204	+0%
POST Baseline	1,193	-	1,208	-1%
POST With Sentry	514	43%	526	-2%
POST With Sentry (error only)	1,068	90%	1,069	-0%
MYSQL Baseline	3,403	-	3,350	+2%
MYSQL With Sentry	511	15%	438	+17%
MYSQL With Sentry (error only)	2,810	83%	2,715	+3%

View base workflow run

onurtemizkan · 2025-10-09T15:43:52Z

@chargome - I updated the PR with edge case handling + E2E tests

This resolves the issue that occurs when an extra `navigation` transaction is created after a prematurely ended `pageload` transaction in React Router lazy routes. This apparently occurs when there's a long-running pageload with lazy-routes (after fetching assets, there are multiple potentially long-running API calls happening). This causes the `pageload` transaction to prematurely end, even before the fully parameterized transaction name is resolved. The reason is that there can be a `POP` event emitted, which we subscribe to create a `navigation` transaction. This ends the ongoing `pageload` transaction before its name is updated with a resolved parameterized route path, and starts a `navigation` transaction, which contains the remaining spans that were supposed to be a part of the `pageload` transaction. This fix makes sure the initial `POP` events are not necessarily treated as `navigation` pointers, which should fix both: - Duplicate / extra `navigation` transactions having a part of `pageload` spans. - Remaining wildcards in the `pageload` transaction names

…es (#19086) This PR will resolve the core reason for the series of fixes / handling for automatic lazy-route resolution for a while. Related: #18898, #18881, #18346, #18155, #18098, #17962, #17867, #17438, #17277 The core issue we have been trying to tackle is not having access to the complete route hierarchy when asynchronously loaded lazy routes are used. React Router provides a route manifest that we can use while matching parameterized transaction names with routes in all cases except this lazy-routes pattern. This problem has been discussed on React Router: - remix-run/react-router#11113 While this has been [addressed](remix-run/react-router#11626) for Remix / React Router (Framework Mode), it's still not available in Library Mode. The manifest contains the lazily-loaded route, only when it's navigated to. While waiting for navigation, our transactions can be dropped for several reasons, such as user behaviour like switching tabs (`document.hidden` guard), hitting timeouts like `idleTimeout`, and potentially other reasons. This results in incomplete transaction naming with leftover wildcards, which caused broken aggregation on the Sentry dashboard. The series of attempts to fix this while keeping automatic route discovery has been prone to race conditions and required special-case handling of each edge case scenario, also requiring a considerable amount of internal logic, affecting our readability and performance. At the end, all failed in giving completely robust and deterministic results on the customers' side. This PR proposes a new option: `lazyRouteManifest` specifically for lazy routes. This will let us have initial information about the route hierarchy. So we can assign correct parameterized transaction names without needing to wait for navigated state. It's a static array of routes in parameterized format (needs to be maintained by the users on route hierarchy updates) like: ```ts Sentry.reactRouterV7BrowserTracingIntegration({ // ... enableAsyncRouteHandlers: true lazyRouteManifest: [ '/', '/pricing', '/features', '/login', '/signup', '/forgot-password', '/reset-password/:token', '/org/:orgSlug', '/org/:orgSlug/dashboard', '/org/:orgSlug/projects', '/org/:orgSlug/projects/:projectId', '/org/:orgSlug/projects/:projectId/settings', '/org/:orgSlug/projects/:projectId/issues', '/org/:orgSlug/projects/:projectId/issues/:issueId', '/org/:orgSlug/team', '/org/:orgSlug/team/:memberId', '/org/:orgSlug/settings', '/org/:orgSlug/billing', '/admin', '/admin/users', '/admin/users/:userId', '/admin/orgs', '/admin/orgs/:orgId', ], }) ``` - This will only be active when `enableAsyncRouteHandlers` is set to `true` - To match URLs with given routes, we mimic React Router's own implementation. - When this is not provided or fails, it falls back to the current behaviour - This manifest is primarily for lazy routes, but the users can also add their non-lazy routes here for convenience or consistency. - Also added E2E tests that will fail when (if at some point) React Router manifests include the lazy routes before navigation, so we'll be aware and plan depending on that manifest instead. - We can do a cleanup for the race-condition / edge-case handling part of the code in a follow-up PR. Closes#19090 (added automatically)

onurtemizkan force-pushed the onur/react-router-long-running-pageload-guard branch from c5b63b9 to b3027fdCompare October 6, 2025 11:21

onurtemizkan marked this pull request as ready for review October 6, 2025 12:14

onurtemizkan requested review from chargome and s1gr1d October 6, 2025 12:14

This comment was marked as outdated.
Sign in to view

chargome approved these changes Oct 8, 2025
View reviewed changes

This comment was marked as outdated.
Sign in to view

onurtemizkan force-pushed the onur/react-router-long-running-pageload-guard branch from 68e2702 to 5608b63Compare October 9, 2025 13:14

s1gr1d approved these changes Oct 9, 2025
View reviewed changes

onurtemizkan added 8 commits October 15, 2025 12:37

fix(react): Add POP guard for long-running pageload spans
249062e

Improve
af57095

Make memory router consistent
52740b9

Improve the loading state checks
9ac974d

Add E2E tests
1ba305f

Stop relying on idle state
8576462

Make it more defensive
851b1b0

Address potential race condition
1fb8a6d

onurtemizkan force-pushed the onur/react-router-long-running-pageload-guard branch from ec0a205 to 1fb8a6dCompare October 15, 2025 11:37

chargome approved these changes Oct 15, 2025
View reviewed changes

chargome self-assigned this Oct 15, 2025

chargome merged commit fc64c47 into developOct 15, 2025
91 checks passed

chargome deleted the onur/react-router-long-running-pageload-guard branch October 15, 2025 11:51

This was referenced Oct 15, 2025
Bump @sentry/react from 8.55.0 to 10.20.0 Pouria-FOSSA/dependabot-metadata-test#71
Closed
Bump @sentry/react from 8.55.0 to 10.21.0 Pouria-FOSSA/dependabot-metadata-test#72
Closed

onurtemizkan mentioned this pull request Jan 29, 2026
feat(react): Add lazyRouteManifest option to resolve lazy-route names #19086
Merged

github-actionsBot mentioned this pull request Jan 29, 2026
feat(react): Add lazyRouteManifest option to resolve lazy-route names #19090
Closed

onurtemizkan mentioned this pull request Feb 23, 2026
ref(react): Remove lazy route machinery in favour of lazyRouteManifest#19466
Closed