chore(release): give tests/migrate its own race shard so the 0.10.0 release workflow can pass

[devin-ai-integration]

What and why

Hotfix from main for the 0.10.0 release. Since #1033 every CircleCI develop build has failed in Go race tests: on the 4-vCPU rest node tests/migrate under -race went from 1370 s to past the 30-minute package timeout (panic: test timed out after 30m0s, FAIL github.com/Open-MBEE/OpenSysML/tests/migrate 1800.098s). main now carries #1033 through #1064, and every publish job in the release workflow requires Go race tests, so a rehearsal or a v0.10.0 tag fails the same way; the nightly, which only publishes green CircleCI commits, has been stalled as well.

• Brings in ci(test): give tests/migrate its own race shard #1060: tests/migrate becomes a race shard of its own (scripts/race-shard.sh migrate, make test-shard SHARD=migrate, make coverage-shard SHARD=migrate); CircleCI go-race-test and go-coverage go from four to five executors, pr.yml from five to six race shards. Its changelog fragment is folded into the 0.10.0 ### Changed section.
• Gives the CircleCI migrate shard the Makefile's 55-minute per-package timeout (make test-shard already uses 55 m). The package alone, with nothing beside it, takes 1423 s under -race on an idle 8-vCPU machine, so on the 4-vCPU large class a node of its own leaves no margin under 30 m.
• Trims the pr.yml shard comment to two lines.

No test or gate changes: the same packages run under the same flags, one more job wide.

How it was verified

• circleci config validate .circleci/config.yml passes; scripts/race-shard.sh migrate prints only tests/migrate and rest no longer lists it (119 packages).
• go test -race -pgo=off -count=1 ./tests/migrate alone: ok 1422.985s (8 vCPU).
• python3 scripts/changelog.py check, python3 scripts/check-doc-ids.py pass; exactly one ## 0.10.0 section.
• The race run itself is CI's: this PR's Go race tests (migrate) / (rest) jobs, then the CircleCI release_rehearsal on main after merge, are the proof.

Checklist

• make test and make lint pass locally (CI configuration change; the Go suite is unchanged)
• Tests added or updated for the change (n/a: CI configuration)
• Documentation extended where it already covers the surface (AGENTS.md shard list, Makefile help text, via ci(test): give tests/migrate its own race shard #1060)
• Changelog: fragment folded into the existing 0.10.0 section (release preparation, per docs/project/releasing.md)
• baselines regenerated and make docs-counts run if a gate count moved (n/a)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/3db6a090aa214667b2a66d372c6e36b1
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/3db6a090aa214667b2a66d372c6e36b1?variant=devin
Requested by: @HuiJun

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

Devin Review found 1 potential issue.

[devin-ai-integration]

🟡 Migration shard retains shorter silence limit

When tests/migrate exceeds 30 minutes, the 55-minute timeout cannot prevent CircleCI from terminating the silent command. go test buffers package output, so observed 32-minute runs still fail the release workflow.

Learn more

CircleCI stops a run step when it produces no output for no_output_timeout. The repository already documents that go test buffers a package's -v output until that package finishes in the solver-absent gate at .circleci/config.yml. This shard contains only tests/migrate, so a package run can remain silent for its entire 23–32 minute duration. Raising Go's test timeout to 55 minutes therefore leaves CircleCI's independent 30-minute silence limit as the effective cap.

Example: A loaded runner completes tests/migrate in 32 minutes. CircleCI sees no output after the initial shard message, kills the step at 30 minutes, and never reaches Go's 55-minute timeout.

Recommended fix: Raise this run step's no_output_timeout above the migration shard's 55-minute Go timeout, or stream periodic output without masking the test exit status.

---

Was this helpful? React with 👍 or 👎 to provide feedback.

[devin-ai-integration]

Not a bug: go test -v streams a single package's output as it runs, so the silence limit is never reached. Evidence from the first release build with this change (pipeline 3995 on main): the migrate node ran 33.2 minutes under no_output_timeout: 30m and passed (ok github.com/Open-MBEE/OpenSysML/tests/migrate 1923.117s). The buffering noted on the REPL gate step applies to multi-package runs; the migrate shard is one package.