Issue: #37207
This issue usually occurs when you enable log backup during a full data import and afterward perform a PITR to restore data at a time point during the data import.
Specifically, there is a probability that this issue occurs if there are a large number of hotspot writes for a long time (such as 24 hours) and if the OPS of each TiKV node is larger than 50k/s (you can view the metrics in Grafana: TiKV-Details → Backup Log → Handle Event Rate).
It is recommended that you perform a snapshot backup after the data import and perform PITR based on this snapshot backup.