[PingCAP docs, Error Codes and Troubleshooting]: Generally this is slow or busy disk on PD, or a network failure between TiDB and PD. Check the PD server state, monitoring data and logs. On TiDB Cloud this is infrastructure-side, so if the TiDB console shows a healthy cluster, it is transient and retryable; persistent 9001s are worth a support ticket with the time window.

Context: TiDB Cloud queries fail with ERROR 9001 (HY000): PD server timeout; start timestamp may fall behind safe point. A TiDB background worker continuously queries the safepoint from PD and reports this error when it fails for 100 seconds.