A model response cache serves one tenant answer to another tenant after a key collision. What is the essential correction?

1. Increase the cache TTL to reduce the number of collisions per day.
2. Remove the tenant dimension entirely to maximize hit rate.
3. Scope cache keys and reuse to the authorized tenant and data context.
4. Hide the answer source while leaving the private content unchanged.