When changing error handling, add/maintain tests that:

• Cover all supported configuration shapes (e.g., useErrorBoundary: true and useErrorBoundary: (err) => boolean).
• Assert the exact outcome of error handling logic (e.g., whether an error is redacted during dehydration/rehydration).
• Keep tests deterministic: don’t manipulate NODE_ENV or other globals unless the behavior truly differs.
• Ensure the test name/expectations match the real trigger (e.g., the hook that throws and causes the boundary to render).
• Use the correct async error pattern so the error is observed by the code path under test (prefer Promise.reject(err) / then(() => Promise.reject(err))).

Example (boundary + config variant):

it('lets errors fall through when useErrorBoundary is falsey', async () => {
  const key = queryKey()

  function Page() {
    useQueries({
      queries: [{
        queryKey: key,
        queryFn: () => Promise.reject(new Error('boom')),
        useErrorBoundary: (err) => err.message.includes('boom'),
      }],
    })

    return null
  }

  // assert: boundary behavior matches the function result
})

Example (redaction rule):

const queryClient = createQueryClient({
  defaultOptions: {
    dehydrate: {
      shouldDehydrateQuery: () => true,
      shouldRedactErrors: () => false,
    },
  },
})

await expect(
  dehydrate(queryClient).queries[0]!.promise,
).resolves.toBeInstanceOf(Promise) // and then assert the thrown error is the original

The goal is confidence that error propagation/redaction behaves exactly as intended across supported modes, without relying on brittle or misleading test setup.