flush data meet error

## Bug Report

### What did you do?
During the mixed deployment of TiDB, PD, and TiKV, we encountered some issues when using BR for log backup, resulting in data being stored under the log_backup_tmp path. We later discovered that the storage was almost full, which triggered the error at the time. However, after we deleted the logs and freed up space, one of the PD nodes (non-leader) continued to frequently refresh the same error log.

### What did you expect to see?
stop error log refresh

### What did you see instead?
the error start time:

the log frequently refresh:

### What version of PD are you using (`pd-server -V`)?
v8.5.3

Start with PD server behavior in the mixed TiDB/PD/TiKV deployment and inspect the error shown in the attached logs after storage under log_backup_tmp fills the disk. Reproduce the disk-full condition, delete the logs and free space, then verify that the non-leader PD node no longer refreshes the same error.

Written by the indexing model from the issue text.