fix(clusters): keep the chosen region when a second region is removed on a single-region tier

[dawsontoth]

⊙ Problem

On a tier that allows only one region (the free tier, which carries allowedRegionIds), removing the second region of a new cluster threw away the first region too: it snapped back to the default US region and its smallest latency (#1275). The free tier's own validation error on the second row ("You can only select one region with this performance tier!") is what asks the user to remove it, so the natural fix-up step is the one that loses their choice.

Reproduced on stage in a real browser before this rebase (Playwright against a local dev server with mocked API responses): pick Medium, choose Europe 45ms for region 1, add a region, switch to Free, remove region 2. Region 1 read US / 90ms, small distribution afterwards. #1780 keeps that US-first correction, so the bug is still there on its branch: with the fix reverted on top of #1780, the two #1275 tests below go red.

The cause is autoSelectRegionBasedOnAllowedRegionIds in ClusterForm.tsx. It corrects the only remaining region whenever that region is not one the tier allows, and it always corrected to US. Removing the second row is what drops the count to one and fires it. The triage suspect, autoPickLatencyDescription in RegionFormInputs.tsx (which #1780 removes), was not involved: rows are keyed by field.id, and removing a row leaves the other row's region and latency untouched on tiers without a region allow-list (pinned by a test below).

Stack: #1780 ← this PR ← #1849 ← #1851 ← #1870. This branch is based on #1780's branch (david/1778-custom-regions), so it lands on top of #1780 instead of #1780 rebasing around it; retarget to stage once #1780 merges. On 2026-10-13 this branch was rebased with #1780 onto stage, after #1917 merged. stage had gained its own ClusterForm.test.tsx (#1846's refused-edit cases), so this branch's region cases now sit beside it in ClusterForm.regions.test.tsx, following the repo's Component.topic.test.tsx convention. Both files keep every case.

❓ Your call: the requirement as filed asks that the first region "remains". It now keeps its region and moves only the latency when the tier does not offer the chosen latency (Europe 45ms becomes Europe 60ms on the free tier), because the alternative is leaving an invalid selection with an error under it. The latency changes without a notice. If you would rather the form never auto-correct after a removal and let the user fix the flagged latency by hand, or show a hint when it does, either is a small change in the same effect.

💡 Solution

The correction now prefers an allowed entry in the region the user already chose, and falls back to US, then the first allowed region, only when the tier offers nothing there.

How it fits #1780: #1780 stores each region plan as a region id, so the effect looks up the surviving id in #1780's regionLookup and matches the catalog region's name (ResolvedRegion.name, the catalog row's region). A custom region (oreg-…) is never corrected, because #1780's effect already skips those ids: the allow-list describes the catalog. Before the rebase the same preference read the entry's regionName directly.

When the tier allows several latencies in the chosen region, the correction takes the first allowed one in catalog order rather than the one closest to the old latency. Today's single-region allow-lists offer one latency per region, so the two can't differ yet.

🔧 Changes

• src/features/clusters/upsert/ClusterForm.tsx — the change in autoSelectRegionBasedOnAllowedRegionIds: the surviving id's catalog name, then a find on it, run before the existing US and first-allowed fallbacks. Nothing else in the effect, including when it runs, changed.
• src/features/clusters/DESIGN.md — records when the correction runs (including on removal), that a custom region is never corrected, and the same-region-first rule.

✅ Verification

Route: a component test driving the real ClusterForm (react-hook-form, zod resolver, Radix selects, #1780's region lookup, the real Remove button) in jsdom.

• src/features/clusters/upsert/ClusterForm.regions.test.tsx (new) — five cases, reading each row as it is shown, because the form now stores a region id: removing a row on an unrestricted tier leaves the survivor untouched; on the free tier the survivor keeps its region whether row 2 or row 1 was removed ([Create Cluster] Deleting the second region will also revert the first chosen region and latency #1275); an already-allowed survivor is untouched; and a survivor in a region the tier does not offer still falls back to US. The fixtures now give each row a region id, and the harness passes feat(clusters): org-owned custom regions — picker group and units on the cluster form #1780's new ClusterForm props (no custom regions).
• Fails on base: on top of feat(clusters): org-owned custom regions — picker group and units on the cluster form #1780, with the same-region find disabled, the two [Create Cluster] Deleting the second region will also revert the first chosen region and latency #1275 cases go red (expected [ 'US / 90ms, small distribution' ] to deeply equal [ 'Europe / 60ms, small distribution' ]) and the other three stay green.
• npx vitest run src/features/clusters/upsert/ClusterForm.regions.test.tsx → 5 passed (exit 0). Full gate via the pre-commit hook at d2c4490: vitest 399 files / 3698 tests passed, oxlint clean, dprint check --staged clean, commitlint ok; npx tsc -b exit 0.
• After the rebase onto stage (d9703d2): npx tsc -b exit 0, and npx vitest run src/features/clusters 203 passed. The full suite passed at the stack's top tree: 5,552 tests, exit 0.
• Not re-run in a browser after the rebase: the Playwright repro above ran on stage before it.

Cross-model review on the rebased branch (one round, vs david/1778-custom-regions): codex (graded), gemini and the Harper domain adjudicator all ran. The adjudicator kept two nits. Taken: the test helper's comment narrated its DOM walk, so it now says only why rows are read as shown. Rejected: explicit import extensions (this repo's imports are extensionless throughout). It dropped gemini's three majors on evidence: ResolvedRegion.name is the catalog row's region (lib/regionLookup.ts:53), the includes call only runs once allowedRegionIds?.length is truthy, and submit re-resolves each id from the same form state the effect writes. Its two open decisions (first allowed latency vs closest, silent latency change) are answered above.

Closes #1275

_{🤖 Generated by Anthropic Claude Code (Claude Opus 5.5); posted via @dawsontoth.}

🤖 Generated with Claude Code

Related PRs: #1780 overlaps (this branch is stacked on it; it rewrote the same correction effect for region ids), #1849 overlaps (stacked on this branch), #1846 overlaps (it created stage's ClusterForm.test.tsx; this branch's cases sit beside it in ClusterForm.regions.test.tsx), 3 others independent
Complexity: easy

_{Review-Coverage: authored=claude; ran=gemini,codex; adjudicated=domain; declined=cursor-grok,cursor-composer,cursor-kimi,cursor-muse; rounds=1; full=1 @ d2c4490}

_{Review-Attention: read ~3m (decisions: same-location-picks-first-allowed-latency, silent-latency-correction) @ d2c4490}

[gemini-code-assist]

Code Review

This pull request updates the region correction logic in ClusterForm to prioritize keeping the user's chosen region when correcting regions on a tier that limits allowed regions, falling back to 'US' or the first available region only if the chosen region is not offered. It also adds corresponding documentation in DESIGN.md and a comprehensive suite of unit tests in ClusterForm.test.tsx to verify this behavior. There are no review comments to address.

[github-actions]

Coverage Report

Status	Category	Percentage	Covered / Total
🔵	Lines	79.17%	14687 / 18550
🔵	Statements	79.1%	15583 / 19698
🔵	Functions	74.56%	3916 / 5252
🔵	Branches	73.68%	11501 / 15608

File Coverage

File	Stmts	Branches	Functions	Lines	Uncovered Lines
Changed Files
src/features/clusters/upsert/ClusterForm.tsx	64.7%	43.94%	77.77%	64.1%	135-144, 149-153, 164-171, 177-181, 184-210, 224-228, 233-237, 285, 288-301, 305, 308, 365, 387, 398, 445-478, 488-495, 498-519, 532-562, 582-583, 589, 593

Generated in workflow #2361 for commit d9703d2 by the Vitest Coverage Report Action