Triage and repair failing CI / CodeRabbit changes on Carson PRs

Triaged 17 open PRs: 16 have failing CI and/or CodeRabbit requests. Beginning parallel repairs on the smallest isolated fixes first (PRs #13685, #13687, #13690, #13692).

PR #13692 triage complete: CodeRabbit duplicate-mock finding is not reproducible in current branch; no code change made. The failed Unit Tests run timed out in apps/dashboard after 600s and did not report a CRM test failure. Moving this PR behind active repair work while continuing other fixes.

PR #13687 fixed and pushed: 32362f60127 clears stale org deals immediately on orgId change; lint, typecheck, and targeted pagination tests pass.

PR #13685 verified: CodeRabbit’s two requested fixes already landed in 29969c0. Relay lint/typecheck and full relay test suite (139 tests) pass; no further commit needed.

PR #13698 fixed and pushed: f8046bcc hardens RBM write paths, clears relay knip, adds P2002 conflict translations/tests, shares admin access check, and types mutations. Relay lint/typecheck/knip pass; new CI queued.

PR #13690 fixed and pushed: 8bcf518 bounds logout cleanup, clears local auth before logout, reliably redirects and resets cleanup state; adds stalled-logout coverage. Relay tests (168), lint, typecheck, and build pass.

PR #13693 fixed and pushed: fd059e8 adds dedupe retry, malformed-email rejection, 10-item contact-method limits, generated Relay mutation types, and focused service tests. Relay codegen/typecheck/lint/targeted test pass; identity full checks blocked by pre-existing workspace artifact issue.

PR #13697 fixed and pushed: 0ecbabf ends stuck status-ID loading on error and fixes CRM formatter CI; CRM lint/tests/build and Relay lint/targeted tests pass. Sales-status sort properly waived pending a cross-subgraph sort contract; full ID-materialization needs architecture follow-up, not an unsafe cap.

PR #13691 confirmed clean: CodeRabbit approved; Unit Tests failure is an unrelated apps/dashboard Vitest coverage timeout (exit 124), not a changed-path test failure. No code change or no-op commit needed.

Disk space unblocked by clearing old /tmp/pr- directories. Resuming triage and repair of failing CI / CodeRabbit changes on Carson PRs.

PR #13703 fixed and pushed: adds focus-visible for accessibility, missing __resolveReference, and syncs supergraph schema. Awaiting new CI run and CodeRabbit review completion. Moving to PR #13694.

PR #13694 re-triaged against live GitHub review threads: four unresolved frontend CodeRabbit issues remain plus an uncommitted backend repository safety refinement. A dedicated repair agent is applying the complete set, running lint, committing, and pushing to the existing branch.

PR #13694 repair committed and pushed as 1e5ad0e4d91c57d878b9e50f69832ce9cb75314e. New CI and CodeRabbit review are running. Local relay lint could not run because the workspace node_modules state file is missing; no install was performed.

Subagent fix_pr_13694_ci completed. Fixes pushed for failing build-docker and test checks. CI is re-running.

Verified PR #13694 CI is now fully green. Fixes pushed by the subagent were successful.

PR #13697: applied final CodeRabbit nitpick comment fix in apps/relay/lib/org-list.ts. Committed and pushed to carson/orgs-status-server-filter as fb760f5bf40c974fd13ba8ef59acb59434b109f9. Local relay biome lint green. CI diagnosis: Unit Tests (dashboard) and Shadow path-filter ↔ turbo-affected are both pre-existing infrastructure flakiness unrelated to this PR (dashboard vitest timeout; edges-agents DTS resolution failure in unrelated packages). active-ci-watch.env created at $HOME/active-ci-watch.env.

PR #13697 CI monitoring update: latest push fb760f5bf40 is green on all relay-relevant and diff-relevant checks (affected, lint, relay test, knip, graphql, build-docker, apollo checks, Shadow path-filter). The only remaining pending checks are the dashboard Unit Tests job (run 30154276218 job 89669672094) and deploy / Deploy relay, both blocked on a dashboard workflow that does not touch this PR. Unit Tests has been IN_PROGRESS for >80 minutes with no logs available via gh while in progress, indicating a GitHub Actions runner/status hang rather than a PR-caused failure. PR diff is comment-only in apps/relay/lib/org-list.ts.

Fresh 2026-07-26 CI triage: PR #13697 has two genuine red checks. Shadow path-filter fails because @texturehq/edges-agents begins its DTS build before @texturehq/edges emits declarations (TS7016). Unit Tests runs successfully long enough to upload coverage but exceeds its 600s shell timeout and consequently lacks lcov.info. Investigating the build graph race first; no Carson DM.

Pushed CI race fix 41e96a26364 to PR #13697: explicit @texturehq/edges-agents#build turbo dependencies serialize DTS output. Local package lint and scoped turbo build passed. Active CI watch recorded; monitoring the new run. Dashboard Unit Tests timeout remains a separate issue.

CI re-check: the DTS race is fixed — Shadow path-filter is green. A separate Unit Tests failure appeared in the Test auth workflow while a dashboard Unit Tests job and auth E2E job are still running. Investigating its exact failed step; no DM.

PR #13697 follow-up fix pushed e09445598d5. Root cause refined: test-auth setup runs root postinstall/build:workspace-packages on cache miss; @texturehq/edges-explore has build:deps that directly rebuilds @texturehq/edges and can clean edges/dist after Turbo considers @texturehq/edges#build complete. Added @texturehq/edges-explore#build to @texturehq/edges-agents#build dependencies. Verification: turbo.json JSON parse, edges-agents lint, scoped @texturehq/edges-agents forced build, and full packages/* forced build all pass locally. Monitoring CI.

PR #13697 CI re-check: all newly triggered checks for 41e96a26364 are pending; no new failures visible. Active CI watch remains in place; monitoring continues silently.

checkpoint saved

checkpoint saved

Fresh verification: Fleet GitHub auth works. PR #13697 has one genuine red Unit Tests job: it hangs after beginning DeviceDetailPage.test.tsx and is killed by the workflow 600s shell timeout; coverage/lcov.info is therefore absent. Investigating the test/module-load hang before proposing a fix; no DM.

Targeted reproduction does not confirm commit 16383afc as the 600s hang source: its DeviceDetailPage test finishes in 2.48s, but fails 73 tests because the edges mock lacks useNotice. The skipped state exits in 587ms. Branch left unchanged; correlating the failing run SHA with current PR head before selecting a repair.

PR #13697 timeout revalidated from the failed CI log: Vitest completed only two files plus the skipped DeviceDetailPage suite, then produced no further output before the 600s timeout; lcov was never finalized. The prior unskip commit is on a separate branch and is not a valid fix (current targeted run fails immediately on a stale useDeleteDevice mock). Running an isolated reproduction/root-cause pass before changing Carson’s branch; no push and no DM.

Sub-agent investigation reached its execution limit without a root-cause conclusion. Fresh CI verification confirms PR #13697 remains blocked only by dashboard Unit Tests timing out at 600s; restarting a focused reproduction/fix pass on the clean PR-head worktree. No push or Carson DM.

Root cause isolated: Vitest stalls transforming the @phosphor-icons/react barrel import in dashboard gridGlyphs.tsx. A focused fix replaces it with direct CSR icon imports; targeted CI-shape suites now complete (158 passed / 73 skipped in 23.91s). Verifying the exact diff and preparing the PR-head commit/push.

Pushed dashboard timeout fix 752ddb5e839 to PR #13697. Replaced the @phosphor-icons/react barrel import in gridGlyphs.tsx with direct CSR imports, preventing Vitest transform stalls. Verified dashboard lint passes with three existing warnings only; CI-shape reproduction finishes in 23.96s (158 passed, 73 skipped). Root-level lint remains blocked by unrelated pre-existing errors in apps/programs and docs. Active CI watch recorded; monitoring new run.

New CI is materially improved: dashboard Unit Tests passed in 2m49s, confirming the Vitest timeout fix. DeepSource coverage now flags 7 introduced uncovered lines in gridGlyphs.tsx because the newly executed module is uncovered; assigning focused branch coverage tests. No Carson DM.

Pushed 5ea6b345434 to PR #13697: focused gridGlyphs coverage addresses the DeepSource uncovered lines. Targeted Vitest coverage and changed-file Biome passed; dashboard lint has only 3 unrelated pre-existing warnings. Active CI watch is running; no DM.