ci(python): run round-trip sweeps on latest only, move experimental legs to their own workflow (#1732)

[JarryShaw]

Please follow the guide below

• You will be asked some questions, please read them carefully and answer honestly

• Put an x into all the boxes [ ] relevant to your pull request (like that [x])

• Use Preview tab to see how your pull request will actually look like

• Searched for similar pull requests

• Followed the coding style (make pylint, make mypy, make isort)

• make test passes, and a test case covers the change. Narrow runs only, listed below.

• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible: N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

Tick the commit type your subject line carries.

• fix — corrects a defect
• feat — adds a feature
• perf — changes performance, not behaviour
• refactor — changes neither behaviour nor performance
• test — tests only
• docs — documentation only
• ci — workflows or build tooling
• release — bumps the version or rolls up a distribution
• chore — anything else

---

Closes #1732. This applies the owner's three rulings on that issue.

• Sweeps on latest only: every other test leg skips the 41 *roundtrip* modules (303 tests). Every other integration leg drops its three *roundtrip*_runtime.py modules (251 tests become 239).
• Option 2: 3.14t and GraalPy move to .github/workflows/experimental.yml ("Experimental Tests"). It reports separately and holds up nothing: no needs:, no workflow_call, and no workflow_run names it, so Unit Tests and its four consumers complete without it. Check names are unchanged. Caps are 140 for the job and 5/10/120 for the steps, with job-level continue-on-error. Ruleset 23497679 requires none of these checks.
• Timing report, option A: util/ci_timing.py reports H1, H2 and H4 on non-blocking legs as warnings. Those are experimental.yml's legs, (KTLO) legs, and (experimental) legs from before the split. Blocking legs stay HARD.
• Docs: workflows.rst now counts thirteen workflows, with a row and a graph node for the new one, and line citations are renumbered.
• Tests: test_sweeps_on_latest_only_1732.py, test_workflows_at_a_glance_1732.py (pins the count, table and graph to the files) and test_ci_timing.py. Each new assertion fails against main.
• Not fixed here: GraalPy's lost runners (ci(python): the GraalPy experimental leg loses its runner 18–25 minutes into its tests on every recent run #1740) and the KTLO half of ci(unit-tests): a KTLO or experimental leg that fails at job level, such as a lost runner, still fails Required checks passed #1738.

[JarryShaw]

Verdict: GOOD TO GO at 69d085079. A Sonnet cross-review checked each of five rounds (author: Opus). Rounds 1–4 each raised findings, all fixed in later rounds.

• Nothing waits on Experimental Tests:
  • every .github/workflows/*.yml was parsed, and no needs:, workflow_run or workflow_call names it;
  • ruleset 23497679 requires only Required checks passed and Compat 3.11–3.15.
• Sweep subset: the unit tier collects 4501 tests on latest and 4198 on the others. Integration collects 251 and 239, and the 12 dropped tests all belong to the three runtime modules.
• Timing report, simulated through _report(): legacy and new experimental legs, and KTLO legs, past their share exit 0 with a ::warning. A blocking Python 3.14 leg still exits 1.
• New assertions fail on main and pass here.
• My own re-run in the PR's tree: the 1732, ci_timing, at-a-glance, 1660 and 1672 test files all pass.

[github-actions]

Coverage: 89.51% (unit tier, Python 3.15, 69d085079, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	3	20	3	95.16%
pcapkit/const	18860	954	2368	868	91.06%
pcapkit/corekit	2455	69	780	35	96.54%
pcapkit/dumpkit	275	3	100	3	98.40%
pcapkit/foundation	2906	164	1016	34	93.57%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17172	225	4588	190	98.07%
pcapkit/toolkit	865	198	298	6	74.38%
pcapkit/utilities	436	2	140	4	98.96%
pcapkit/vendor	4417	2342	1010	157	43.38%

Per-file detail: the coverage-html artifact of this run.