Claude Opus 4.8 Review

Which disables can be dropped

1. Both useIsHydrated copies (packages/x-charts/src/hooks/useIsHydrated.ts, packages/x-chat-headless/src/chat/internals/useIsHydrated.ts) -- avoidable now, low risk. The rule fires only on passive useEffect. Two options:

   • Lowest friction (no new dep): swap React.useEffect -> useEnhancedEffect. It is already a dep in both packages, is useLayoutEffect on the client / SSR-safe, is not flagged by the rule, and it matches the gaearon gist the file comment already cites ("Option 2: useLayoutEffect"). Bonus: removes the one-frame unhydrated flash.
   • More idiomatic: mirror the existing precedent. packages/x-data-grid/src/hooks/utils/useIsSSR.ts:7 already does exactly this with useSyncExternalStore (via use-sync-external-store/shim, React-17-safe), no effect at all -- useIsHydrated is just its inverse (clientSnapshot = () => true, serverSnapshot = () => false). This needs adding the shim dep to those two packages, so the useEnhancedEffect swap is the lighter touch.

2. @typescript-eslint/no-shadow: 'off' (eslint.config.mjs, global) -- reconsider. This turns off a real correctness rule across the entire monorepo to unblock a version bump -- the heaviest hammer in the PR. If it is masking a small number of new violations from the parser bump (8.59 -> 8.60), prefer fixing them or scoping the off to the offending files; at minimum add a comment/TODO explaining why.

Leave as-is (legitimate, not worth churning in a bump)

• useMounted defer branch -- intentional post-paint passive effect; the non-defer branch already uses useEnhancedEffect and is not flagged.
• The 5 charts-premium setProgram(...) effects -- the effect owns the GL program lifecycle (cleanup calls dispose()); setState there is appropriate.
• useGridSelector (exhaustive-deps + new use-memo disable) -- hand-optimized hot path passing a module-level EMPTY deps array on purpose; fine.
• set-state-in-effect: 'off' for test files -- fine. For docs -- acceptable (demos), though docs ideally model best practice.

Out of scope here -- track as a follow-up

The ~15 remaining sites all share one shape: a useEffect that mirrors/adjusts a prop into local state (GridEditBooleanCell, GridEdit{Input,LongText,Date}Cell, the GridFilterInput* family, Month/YearCalendar, GridColumnHeaderItem, the GridActionsCell clamp, Toolbar, ToolbarContext, tree-view itemPlugin). The React-blessed fix is set-state-during-render with a stored previous prop -- the same pattern useGridSelector already uses at packages/x-data-grid/src/hooks/utils/useGridSelector.ts:83, which the linter accepts. Each has behavioral nuance (StrictMode double-invoke, meta guards), so they should not be refactored inside a dependency bump. Worth a tracked tech-debt issue.