perf(client-runtime): keep turn and checkpoint refs stable while streaming

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}

[t3dotgg]

Problem

During assistant streaming every text delta arrives as a thread.message-sent event. The reducer rebuilt latestTurn as a new object and mapped checkpoints into a new array even when nothing in them changed. Any selector or memo keyed on thread.latestTurn or thread.checkpoints recomputed on every delta for no reason.

Fix

• thread.message-sent and thread.session-set pass the recomputed latest turn through reuseLatestTurn, which returns the previous object when every field matches.
• rebindCheckpointAssistantMessage returns the input array when no checkpoint for that turn needs a new assistantMessageId.
• The messages array still changes on every delta, as it must.

I checked every latestTurn and checkpoints consumer in apps/web, apps/mobile, and packages/client-runtime. All of them compare values, none rely on a fresh reference.

Tests

• vp test run packages/client-runtime/src/state/threadReducer.test.ts (30 passed, 2 new: refs stay stable across a streaming delta, refs change when the first assistant message binds the turn)
• vp run --filter @t3tools/client-runtime typecheck
• vp lint and vp fmt --check on the two changed files

Change authored by Claude Fable 5.1 running in Claude Code.

---

Note

Low Risk
Pure reducer structural-sharing change with behavioral tests; no API or security surface change, only reference identity when values are equal.

Overview
Reduces unnecessary re-renders during assistant streaming by preserving object identity on thread state when turn and checkpoint data are unchanged.

applyThreadDetailEvent now routes recomputed latestTurn through reuseLatestTurn, which returns the previous object when every field (including optional sourceProposedPlan) matches. rebindCheckpointAssistantMessage skips mapping and returns the existing checkpoints array when the turn’s checkpoint already has the right assistantMessageId. thread.message-sent and thread.session-set use both behaviors; messages still get a new array on each streaming delta as before.

Tests assert reference stability across a streaming text delta and new references when the first assistant message binds assistantMessageId on latestTurn and checkpoints.

^{Reviewed by Cursor Bugbot for commit 9c8173e. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Preserve latestTurn and checkpoints object references during streaming in threadReducer

• Adds reuseLatestTurn helper to applyThreadDetailEvent that returns the prior latestTurn object when all compared fields (turn identity, state, timestamps, assistant message ID, and sourceProposedPlan thread/plan IDs) are equal, and returns the computed value otherwise
• Updates rebindCheckpointAssistantMessage to return the input checkpoints array unchanged when no checkpoint needs rebinding, and only allocates a new array when a binding actually changes
• Routes thread.message-sent and thread.session-set through reuseLatestTurn, preserving existing turn-state and settlement rules while avoiding new object references for equivalent records
• Adds regression tests verifying that streaming deltas mutate message text and collection without replacing latestTurn or checkpoints references, and that the first assistant message correctly binds its ID into both
• Behavioral Change: latestTurn and checkpoints references now remain stable across streaming deltas when content is unchanged; consumers relying on reference inequality to detect any reducer pass will no longer see updates on no-op deltas

^{Macroscope summarized 9c8173e.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.2 KiB	13.3 KiB	+53 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+88 B (+0.2%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.2 KiB	13.3 KiB	+62 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+62 B (+1.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: 98725df · PR result: 9c8173e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 9c8173e

Macroscope's review found this PR approvable — This is a narrowly scoped client-runtime performance optimization that stabilizes existing state references during streaming without changing the underlying thread data or adding a capability. Targeted tests verify both reference reuse and the cases that must still allocate updated state.

^{You can add or adjust custom eligibility rules. Learn more.}