test: complete the h2 settings round-trip before asserting single-session multiplexing

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ostapondo]

The fetch POST bodies dispatch concurrently (#5494) test flakes on the
macOS runners: sessions.size comes out 2 instead of 1, e.g.
https://github.com/nodejs/undici/actions/runs/30984506686

Over h2 the client counts as busy until the server's SETTINGS frame lands.
With h2c nothing covers that window, so on a slow runner the second fetch
goes out while it is still open and the Agent opens a second connection.

I checked with a TCP proxy that delays server to client by 80ms: no warmup
gives sessions=2 every run, a warmup gives sessions=1 every run.

So warm the session up first, like the SSE test above it does. The #5494
check still works, a client that serialises hangs on secondArrived until
the timeout fires.

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Review ready

[ostapondo]

Reopened - closing this was my mistake, the flake is still there. Rebased on main so CodeQL goes green.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.44%. Comparing base (6aef2fb) to head (fe72eef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #5656 +/- ##
==========================================
+ Coverage 93.33% 93.44% +0.10% 
==========================================
Files 110 110 Lines 38723 38723 ==========================================
+ Hits 36143 36184 +41 + Misses 2580 2539 -41 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.