[aglinxinyuan]

What changes were proposed in this PR?

This fixes the random unit-test timeouts on the macOS frontend leg (#9197) at their cause: memory that grows across spec files inside each Vitest worker.

The Angular unit-test builder runs Vitest with isolate: false, so a worker runs all its spec files against one jsdom document and one copy of Vitest. Three kinds of worker-wide reference kept every finished test's TestBed graph alive, and any one of them was enough on its own. On the 7 GB macOS runner that came to about 2.4 GB per worker, and the VM swapped until whichever test was running timed out.

1. Root services now release what they start. These are production bugs too.
   • WorkflowActionService.ngOnDestroy() destroys the shared model.
   • SharedModel.destroy() calls wsProvider.destroy() instead of a conditional disconnect(). This releases the provider's interval, BroadcastChannel subscription and listeners, and stops a provider that never connected from retrying forever.
   • WorkflowWebsocketService.ngOnDestroy() stops the heartbeat and closes the socket.
   • JupyterPanelService.ngOnDestroy() removes its window listener.
2. jsdom's selector cache is emptied between tests. It caches every selector with the nodes it matched, including the #root<N> query TestBed makes for each fixture. The new setup file is src/jsdom-selector-cache-setup.ts.
3. Vitest 4's mock registry no longer keeps every mock alive. @vitest/spy held every vi.fn() / vi.spyOn() in a Set that is never pruned. A Yarn patch backports Vitest 5's WeakRef registry, since Angular 21 still requires Vitest 4. The patch carries Vitest's MIT code, so it is listed in LICENSE, with the license text in licenses-3rd-party-code/vitest.txt.

Measured locally in the macOS shape (2 workers, coverage, full suite):

	main	this PR
Peak footprint per worker	2.9–3.1 GB	1.5–1.9 GB
Peak footprint of the process tree	8.7–9.0 GB	5.6–6.2 GB
Vitest Duration	85–115 s	65–66 s

Notes for review:

• Why not isolate: true? It also bounds memory, but it roughly doubled the Vitest run locally (230 s against about 100 s).
• The spy patch is needed. Without it the process-tree peak went back to 7.4 GB. Clearing or resetting mocks at the end of each file instead reached only 6.8 / 7.0 GB.
• The patch has to be redone on every Vitest bump. The resolutions entry names the exact @vitest/spy version, so it stops applying whenever Vitest is bumped.

Any related issues, documentation, discussions?

Closes #9197. Related: #7713 raised the test timeouts and #7975 capped the worker heap. Both treated the symptom.

How was this PR tested?

• Run before main started requiring Node 24, on this branch's earlier commits:
  • yarn --cwd frontend run test:ci: 244 files passed, 6857 tests passed, 1 skipped.
  • The full suite twice at 2 workers with memory sampled; both runs passed all 244 files. The numbers are in the table above.
• The new tests fail against the old code and pass with the fix:
  • SharedModel destroy: 2 tests
  • WorkflowActionService, WorkflowWebsocketService, JupyterPanelService teardown: 1 test each
• On the current head, which includes the latest main:
  • yarn install --immutable passes.
  • The test build type-checks every spec.
  • prettier-eslint and eslint are clean on the changed files.
  • The test run itself is left to CI, because main now needs Node 24 and my machine has Node 20.

Was this PR authored or co-authored using generative AI tooling?

Generated-by: Claude Code (Claude Opus 5.5)

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

[github-actions]

Backport auto-label report

This fix: PR was checked against each actively-supported release branch. A release/* label nominates a backport target; the branch's release manager approving this PR is what sends the fix there. The required Backport Approvals check stays red until every label below is approved, so each manager either approves or removes their own label — which is why the labels left on a merged PR are exactly the branches it reached.

Release branch	Analysis
✅ release/v1.3	Already labeled — this fix is queued to backport here. @mengw15 decides: approving sends the fix here, removing this label declines it. The merge waits on one or the other.
✅ release/v1.2	Already labeled — this fix is queued to backport here. @xuang7 decides: approving sends the fix here, removing this label declines it. The merge waits on one or the other.

Auto-label run.

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• Contributors with relevant context: @yangzhang75, @mengw15, @zyratlo
  You can notify them by mentioning @yangzhang75, @mengw15, @zyratlo in a comment.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.98%. Comparing base (c57e6b0) to head (6f85798).

Additional details and impacted files

@@            Coverage Diff            @@
##               main    #9198   +/-   ##
=========================================
  Coverage     91.98%   91.98%           
+ Complexity     5084     5083    -1     
=========================================
  Files          1281     1281           
  Lines         55806    55810    +4     
  Branches       7076     7075    -1     
=========================================
+ Hits          51333    51337    +4     
  Misses         2732     2732           
  Partials       1741     1741           

Flag	Coverage Δ		*Carryforward flag
access-control-service	77.44% <ø> (ø)		Carriedforward from b195ae7
agent-service	99.16% <ø> (ø)		Carriedforward from b195ae7
amber	86.42% <ø> (-0.01%)	⬇️	Carriedforward from b195ae7
computing-unit-managing-service	63.52% <ø> (ø)		Carriedforward from b195ae7
config-service	87.50% <ø> (ø)		Carriedforward from b195ae7
file-service	82.29% <ø> (ø)		Carriedforward from b195ae7
frontend	96.53% <100.00%> (+<0.01%)	⬆️
notebook-migration-service	83.77% <ø> (ø)		Carriedforward from b195ae7
pyamber	98.54% <ø> (ø)		Carriedforward from b195ae7
workflow-compiling-service	74.25% <ø> (ø)		Carriedforward from b195ae7

*This pull request uses carry forward flags. Click here to find out more.

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

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

⚠️ Benchmark changes need a look

🟢 0 better · 🔴 10 worse · ⚪ 5 noise (<±5%) · 0 without baseline

Compared against main c57e6b0 benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

	config	throughput	MB/s	latency	max Δ latest / 7d
🔴	bs=10 sw=10 sl=64	449	0.274	21,044/34,134/34,134 us	🔴 +20.2% / 🔴 +147.8%
🔴	bs=100 sw=10 sl=64	935	0.571	104,841/132,479/132,479 us	🔴 +21.1% / 🔴 +42.2%
⚪	bs=1000 sw=10 sl=64	1,154	0.704	863,091/905,924/905,924 us	⚪ within ±5% / 🔴 -11.6%

Baseline details

Latest main c57e6b0 from same runner

config	metric	PR	latest main	7d avg	Δ latest	Δ 7d
bs=10 sw=10 sl=64	throughput	449 tuples/sec	509 tuples/sec	941.58 tuples/sec	-11.8%	-52.3%
bs=10 sw=10 sl=64	MB/s	0.274 MB/s	0.31 MB/s	0.575 MB/s	-11.6%	-52.3%
bs=10 sw=10 sl=64	p50	21,044 us	17,511 us	10,991 us	+20.2%	+91.5%
bs=10 sw=10 sl=64	p95	34,134 us	31,733 us	13,776 us	+7.6%	+147.8%
bs=10 sw=10 sl=64	p99	34,134 us	31,733 us	16,872 us	+7.6%	+102.3%
bs=100 sw=10 sl=64	throughput	935 tuples/sec	1,015 tuples/sec	1,264 tuples/sec	-7.9%	-26.0%
bs=100 sw=10 sl=64	MB/s	0.571 MB/s	0.619 MB/s	0.772 MB/s	-7.8%	-26.0%
bs=100 sw=10 sl=64	p50	104,841 us	97,771 us	86,073 us	+7.2%	+21.8%
bs=100 sw=10 sl=64	p95	132,479 us	109,378 us	93,135 us	+21.1%	+42.2%
bs=100 sw=10 sl=64	p99	132,479 us	109,378 us	111,124 us	+21.1%	+19.2%
bs=1000 sw=10 sl=64	throughput	1,154 tuples/sec	1,169 tuples/sec	1,304 tuples/sec	-1.3%	-11.5%
bs=1000 sw=10 sl=64	MB/s	0.704 MB/s	0.713 MB/s	0.796 MB/s	-1.3%	-11.6%
bs=1000 sw=10 sl=64	p50	863,091 us	854,601 us	841,986 us	+1.0%	+2.5%
bs=1000 sw=10 sl=64	p95	905,924 us	887,282 us	881,107 us	+2.1%	+2.8%
bs=1000 sw=10 sl=64	p99	905,924 us	887,282 us	918,208 us	+2.1%	-1.3%

Raw CSV

config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,445.86,200,128000,449,0.274,21044.39,34133.80,34133.80
1,100,10,64,20,2139.44,2000,1280000,935,0.571,104840.86,132478.58,132478.58
2,1000,10,64,20,17328.36,20000,12800000,1154,0.704,863090.91,905923.67,905923.67