# Corpus rows today's rules do not select a failing class for, and why that is not a selection gap.
# A miss not listed here fails ci-select.py --replay. Format: <10-char head sha> <class>.

# A Darling class failed on a run whose changed files (as last recorded for the pull request) are
# documentation only. The failing attempt ran on an earlier head of the pull request that held code, so the
# recorded list is not the list that run saw.
cc29d98125 Darling.Tests.RepoFileAdoptionTests
7b64070c98 Darling.Tests.DarlingDeployRollbackRetentionTests
e6dbdbea94 Darling.Tests.IncidentAttachmentDeliveryTests

# A Lite class failed on a Darling-only diff, before the Reads=Darling selection (build.yml, 2026-10-04)
# existed or while the pull request still carried the older workflow. The Reads=Darling set is derived from the
# class source (the ci-select.py reads-classes scan since #5961; DarlingReadsTraitGuardTests before it), so a class outside it reads no Darling file: the diff did not cause
# the failure.
d0d482835c PerformanceMonitorLite.Tests.PerformanceCalendarDataTests
cd34bbcb31 PerformanceMonitorLite.Tests.StatusBarSizeReadLockTests
1792370ebd PerformanceMonitorLite.Tests.McpPageContractTests
32239fd349 PerformanceMonitorLite.Tests.MemoryTrendGrantJoinToolTests
ad89ad83a7 PerformanceMonitorLite.Tests.LocalClockBucketKeyTests
08ca596e65 PerformanceMonitorLite.Tests.LocalClockBucketKeyTests
d064b8ebc5 PerformanceMonitorLite.Tests.LocalClockBucketKeyTests
1e9c8ce764 PerformanceMonitorLite.Tests.LocalClockBucketKeyTests
95a42dab2d PerformanceMonitorLite.Tests.LocalClockBucketKeyTests
5219f29479 PerformanceMonitorLite.Tests.AnomalyDetectorTileBehaviourTests
e4cb4a3699 PerformanceMonitorLite.Tests.AnomalyDetectorTileBehaviourTests
1c726d7f67 PerformanceMonitorLite.Tests.AnomalyDetectorTileBehaviourTests
c0adc53226 PerformanceMonitorLite.Tests.ScenarioTests
68ddf17466 PerformanceMonitorLite.Tests.ScenarioTests
a75333e8e2 PerformanceMonitorLite.Tests.ScenarioTests
801cb7c1d8 PerformanceMonitorLite.Tests.ScenarioTests
63ba0d2893 PerformanceMonitorLite.Tests.DuckDbSentinelConnectionTests
411d87afbe PerformanceMonitorLite.Tests.FinOpsTests
eca647c2f4 PerformanceMonitorLite.Tests.QueryStoreRegressionsToolTests
eca647c2f4 PerformanceMonitorLite.Tests.SignificantWaitsToolTests
# A Darling class failed on a pull request whose recorded changes are Lite files only (cut 3a/3c of #5459). The
# class source names no Lite path (the Reads=Lite set is derived from it, by the ci-select.py reads-classes scan since #5961 and by LiteReadsTraitGuardTests before that, and these carry
# none; the one 'Lite' in PlanSync4512CancellationTests is a doc comment), so no Lite edit can reach what it
# tests: each is a timing or live-store test that failed on a first attempt, not a selection gap.
c3a18e1785 Darling.Tests.PgTargetWaitTests
67bf55c45a Darling.Tests.ClipboardTextRetryBudgetTests
48216212d0 Darling.Tests.PlanSync4512CancellationTests

# Map back-test (ci-select.py --replay --map, #5459): a class failed on a run whose changed files do not reach it,
# and the failure is not a selection gap. Each class was checked against the corpus: none of the run's changed
# product files is named by the class, none is covered by it, and the class also failed on runs with a different
# change set, on a push to dev, or as the only failing class of one shard job (a flake or an environment fault).
# The two mass failures of 2026-09-26 and 2026-09-27 (a store migration broke the shared live-store fixture) are not
# here: the map builder now links every class of a big xunit collection to what the collection's fixture ran.
02ae3efb1f Darling.Tests.StatementFilterAlertTests
03d9905807 Darling.Tests.PgServerLogTailRotationLiveTests
0654199a72 Darling.Tests.PgServerLogTailRotationLiveTests
0ac14fc78e Darling.Tests.DarlingAnomalyBaselineTests
0b92d8248f Darling.Tests.PgServerLogTailRotationLiveTests
0c58c0eb56 Darling.Tests.PgTargetSampledWaitLiveTests
0c983ef54d Darling.Tests.PgLogBurstAndRotationLiveTests
11729239b1 Darling.Tests.XminHorizonPersistenceReadTests
2c39e009ff Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
33932296ca Darling.Tests.StoreStatementStatsLiveTests
3af3ade9ea Darling.Tests.StoreStatementStatsLiveTests
3b2d41baae Darling.Tests.QueryStoreIntervalWideGridLiveTests
3bb3ae8fe9 Darling.Tests.TrendPayloadBudgetLiveTests
3de285b652 Darling.Tests.DarlingAnomalyBaselineTests
40284b0061 Darling.Tests.PayloadDimensionLiveTests
4756659045 Darling.Tests.QueryStoreLegacyRowIndexLiveTests
4a3f52389e PerformanceMonitorLite.Tests.TimeHonestyRungReadTests
4f8499a944 PerformanceMonitorLite.Tests.TimeHonestyRungReadTests
6ac72f8605 Darling.Tests.StoreStatementStatsLiveTests
6de10f4dcb Darling.Tests.StoreSessionTimeZoneLiveTests
87836cc97e Darling.Tests.DarlingAnomalyBaselineTests
87e78aa474 PerformanceMonitorLite.Tests.DataStartBannerProbeFailureTests
88204f7cd2 Darling.Tests.RdsDeadlockIngestorTests
8875188337 Darling.Tests.DarlingAnomalyBaselineTests
89c271dd17 Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
8a800ed6c6 PerformanceMonitorLite.Tests.LiteDeltaSeederTests
8b3c5767f0 Darling.Tests.SqlServerStoreTileBehaviourTests
8f95de197e Darling.Tests.StatementFilterAlertTests
93d34a73cd Darling.Tests.RollupBackfillLiveTests
99ad6aebf6 Darling.Tests.PlanSync4512CancellationTests
9be59f1a95 Darling.Tests.CollectionLogDrainForensicsStoreTests
a3b5749937 Darling.Tests.DarlingAnomalyBaselineTests
af31e98db0 Darling.Tests.StoreSessionTimeZoneLiveTests
b33399f219 Darling.Tests.DarlingCloudIdentityProbeTests
c20797c24a Darling.Tests.PgDeadlockNormalizationTests
c9877aad46 Darling.Tests.DarlingAnomalyBaselineTests
c9a1a81456 Darling.Tests.QueryStoreTopLiteralEndStraddleLiveTests
d0d482835c Darling.Tests.XminHorizonPersistenceReadTests
d0fb0ff183 Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
d5370df470 Darling.Tests.DarlingAnomalyBaselineTests
d6c02bf28a Darling.Tests.DarlingStoreUpgradeRevertTests
ed114b5b17 Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
f07bfc441f PerformanceMonitorLite.Tests.TrendEmptyParityToolTests

# The changelog splice (release archive) run whose failing class is a live-database class that no changelog
# change can affect. Since #5614 the three splice files (CHANGELOG.md, docs/changelog/<version>.md and
# tools/changelog/archive-census.txt) are documentation, so a splice takes the docs fast path and selects no
# class; the failure on that run was not caused by the splice.
e20c881360 Darling.Tests.MigrationUpgradeLadderLiveTests

# Map replay against the first nightly map (#5459, run 37892110132): six misses left after the map's own links, each
# checked against the failing job's log (measured) and the corpus. None is a selection gap; none is listed to hide
# one. A sha here is the head of a pull request run, and the map, if it were live, would have given that run no
# reason to run the class because the class is unrelated to the diff.
#
# LiteServerTagsStoreTests (a304350fe3, measured): the diff is the Lite logger (AppLogger, App.xaml.cs, the
# collector runner) and the failure is a TimeoutException from LocalDataService.OpenWriteConnectionAsync, "Could not
# acquire database write lock", on the class's first write and then its second. Nothing in the diff opens or holds
# a store write lock. Every later head of the same pull request (9981020a8a, 54244d4282, f415eae60c) passed, with
# the logger files still in its diff, and no other row of the corpus has this class failing. A shared-fixture lock
# timeout on a loaded runner (the lock holder is inferred, not seen).
a304350fe3 Lite.Tests.LiteServerTagsStoreTests
#
# EventWindowedReadsAreBoundedLivePostgresTests (the four heads of one pull request on 2026-09-25, measured): the
# test had been failing on dev itself from the merge of #4235 (06:11Z) until its fix, #4275 (10:34Z): the corpus
# holds it failing on two dev pushes (f5e001fe65, 659e11a39b) and on nine other pull-request heads that day, all
# changing other things, nearly all with "Darling PG tests (0)" as the failed job. The four heads ran 09:33Z to
# 10:14Z, inside that window; the first one's log shows the plan test's own message ("expected the floor to exclude at
# least 2 of the 4 seeded old-day chunks"). The change the four heads carry (how the store host is set up) does not
# explain the failure: the failure predates it.
f1a1894fd8 Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
69ce7ff7be Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
4b3fe51a97 Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
3747867b5a Darling.Tests.EventWindowedReadsAreBoundedLivePostgresTests
#
# StoreMetricsLatestSkipScanLiveTests (3bb3ae8fe9, measured): attempt 1 of the run failed in "Darling PG tests (2)"
# only, five classes within one second (06:52:04 to 06:52:05), with "57P03: the database system is in recovery
# mode" (the shard's PostgreSQL server had restarted after a backend crash) and the scratch-database drops retrying
# on the same error. Attempt 2 of the same head passed every shard. A first-attempt environment fault of the
# shard's server; the 28 changed files (viewer data-start work) did not cause it.
3bb3ae8fe9 Darling.Tests.StoreMetricsLatestSkipScanLiveTests

# Shared-project replay (#5951): four failures on three pull requests that changed PerformanceMonitor.Common files, found when
# the keep-full rule for the shared projects was narrowed to constants, enums and registration files. None of the four classes has a
# map edge to the shared file the pull request changed, and each failure is not caused by it. LiteDeltaSeederTests (7859d25653) and
# DarlingUpgradeInstallTreeWaitTests (950bb7dbbd) failed on a first attempt and are green on the same sha now.
# LatestValueLookbackLivePostgresTests (1e52a8e34f, 28c0765ab1) failed a plan-shape assertion on PgFactCollector.DatabaseSizeSql,
# a file that pull request never touched.
7859d25653 PerformanceMonitorLite.Tests.LiteDeltaSeederTests
950bb7dbbd Darling.Tests.DarlingUpgradeInstallTreeWaitTests
1e52a8e34f Darling.Tests.LatestValueLookbackLivePostgresTests
28c0765ab1 Darling.Tests.LatestValueLookbackLivePostgresTests
