Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

2. What did you expect to see? (Required)

3. What did you see instead (Required)

In #8721, we introduce a rule that: "if the regions of a given table is not totally removed from TiFlash, skip physically dropping it". However, it does not set the succeeded = false, so after all it will update the last_gc_safepoint.
When the whole cluster has few write commands, the gc_safepoint may not change. It make that the physically dropping only run after the gc_safepoint is changed again. Which will slow down the data disk reclaim.

4. What is your TiFlash version? (Required)