Redis/cache memory growth

A production symptom appears first: flaky tests, stale state, slow UI, late cleanup, or growing memory. This shows up in API gateways, profile caches, feature flag caches, permission lookups, and hot service processes.

Assuming Redis TTL protects a separate in-process cache.

Fix the root scheduling, cleanup, or concurrency pattern instead of adding arbitrary waits.

try {
  await operation();
} finally {
  cleanup();
}

The visual proof is the active runtime event: watch what is retained, blocked, queued, or released at each step.

Reproduce the symptom with a small test, apply the safer pattern, then verify the flaky, slow, or leaking behavior disappears.