[sentry-junior]

Experiment stacked on #129387. Not meant to merge. I'll post timings and then close it.

It includes the ci/acc-exp-poll and ci/acc-exp-chrome-reuse commits and adds one change: XDIST_WORKERS in acceptance.yml goes from 2 to 3.

On x64, 3 workers made no difference (test time −2%, within noise) and jobs were 5% longer. On Arm the runner is only about 67% busy during tests, so a third worker may help here, especially once per-test Chrome churn is gone.

via @joshuarli.

--

View Session [Sentry]

[sentry-junior]

Experiment results (Arm, 5 acceptance shards)

Acceptance test-step seconds per shard, from the GitHub Actions step timestamps:

Run	Per-shard test step (s)	Sum	Slowest step	Job median / slowest	Result	Reruns
Baseline #129387 (2 workers) 41142216797 #2	138, 148, 155, 169, 169	779	169	286 / 308	195 passed, 7 skipped	1 (IssueDetailsTest.test_stripped_event)
#129404 0.1s poll + failure-only debug capture 41145962251	125, 143, 146, 154, 160	728 (−7%)	160	275 / 284	195 passed, 7 skipped	0
#129405 + Chrome reuse 41145975549	105, 118, 119, 126, 146	614 (−21%)	146	246 / 286	195 passed, 7 skipped	1 (ReactAuthTest.test_organization_sso_with_totp)
#129405 + Chrome reuse, typing fix 41146968697	96, 114, 116, 123, 135	584 (−25%)	135	241 / 251	195 passed, 7 skipped	2 (both ReactAuthTest TOTP tests)
#129406 + Chrome reuse + 3 workers 41145986488	87, 100, 106, 118, 121	532 (−32%)	121	237 / 258	195 passed, 7 skipped	1 (ReactAuthTest.test_organization_sso_with_totp)
#129406 + Chrome reuse + 3 workers, restacked 41146990178	102, 107, 109, 115, 115	548 (−30%)	115	237 / 257	195 passed, 7 skipped	1 (AcceptOrganizationInviteTest.test_create_account_and_accept_invite, element not interactable)

Flakiness: all 4 Chrome-reuse runs had a rerun on shard 3, and 3 of them were test_auth_react.py TOTP tests (StopIteration in submit_second_factor: the visible OTP input check passed, but no displayed input was found when the elements were queried again). The runs with a fresh Chrome had 0 TOTP reruns: baseline, #129404, and 6 recent successful master acceptance runs. Chrome reuse probably exposes a race or leaked state there.

Other checks: Backend, backend typing (after the fix on #129405/#129406), getsentry dispatch, and getsentry / backend passed on all three PRs; getsentry / acceptance was skipped.

Cost: each job still spends about 2 minutes on setup, so job time drops less than test time. The job median goes from 286s to 275s (poll), 241–246s (reuse), and 237s (reuse + 3 workers).

This PR (#129406): 3 workers saves another ~7–9% of test time and ~15–25s on the slowest shard compared with reuse alone, but it depends on Chrome reuse and has the same flake risk. Worth revisiting after reuse is deflaked. Closing as an experiment.