fix(admin): invalidate controllerVersion on machineRedeploy and machineUpgrade success

[pandemicsyn]

Summary

Adds an immediate controllerVersion query invalidation to the onSuccess callbacks of both machineRedeploy and machineUpgrade mutations in KiloclawInstanceDetail.tsx.

Previously, the Gateway section could show stale data (e.g. "Unavailable until redeploy") for up to 5 minutes after a redeploy or upgrade succeeded because the controllerVersion query has a 5-minute staleTime and was only re-invalidated once the machine polled back to running state — not at mutation success time.

The fix adds an immediate invalidation at onSuccess (guarded by data?.user_id), clearing the stale cache right away. The existing deferred re-invalidation via the useEffect polling loop (fires once the machine is confirmed running) is preserved alongside this immediate invalidation.

Verification

• Code review only; change is a pure query-cache invalidation with no server-side logic.
• Pattern mirrors the existing controllerVersion invalidation in the useEffect restart-completion hook (lines 1133–1140 in the file).
• No regressions possible — invalidating a query early just causes an extra refetch, not incorrect state.

Visual Changes

N/A

Reviewer Notes

The double-invalidation is intentional: the immediate invalidation at onSuccess kicks out the stale cached value before the machine has restarted, and the useEffect invalidation fires a second time once the machine is confirmed running and the new controllerVersion is actually available. This is the correct two-phase pattern for post-restart UI updates.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 files)

• src/app/admin/components/KiloclawInstances/KiloclawInstanceDetail.tsx

---

_{Reviewed by gpt-5.4-20260305 · 449,029 tokens}