RTK Query Without “Refresh Everything”

At WikiMind, we tuned the cache so it behaves like an adult: we set up 44+ RTK Query tags for targeted invalidation on an enterprise platform with 1200+ DAU.

What did that buy us? When one document changes, we don’t refetch the entire collection tree. Only what’s actually affected gets updated. Less network noise, fewer “why is everything flashing again?”, and more of the feeling that the frontend isn’t still living in 2019.

We also added a Reducer Manager: reducers are added dynamically when the relevant page mounts. In other words, slices are loaded lazily, and the app doesn’t carry unnecessary baggage from the first screen.

Against the backdrop of the constant push for performance and DX, this is basically a must. When a product grows, things like this are no longer “optimizations for later” — they’re how you avoid pain in production over your morning coffee ☕️