# Expected Diagnosis: periodic-tidb-cpu-spike-during-incident-window

This case should conclude that the TiDB CPU spikes on `db-tidb-0` are caused by
TiDB GC being repeatedly triggered under memory pressure, with non-prepared
statement plan cache growth keeping memory above the GC trigger threshold.

The expected answer should make these points explicit:

- the issue is scoped to `db-tidb-0`
- the root cause is TiDB GC under memory pressure, not a generic application spike
- the memory driver is plan cache growth for non-prepared statements
- the evidence chain should mention low query pressure, GC activity, and heap
  profile / plan cache signals
- the answer should not stop at TiProxy load balancing or application-side load
  as the final root cause

What should fail this case:

- stopping at a vague CPU-only explanation
- attributing the issue to an application-side surge without the memory/GC chain
- blaming TiProxy load balancing as the primary root cause