A temporary CPU lockup on the active node causes the node to stop updating the HA table. As a result, the standby node becomes active. However, once the original node recovers, it resumes operation and incorrectly marks itself as active again, causing a split-brain scenario.

Steps to reproduce:
1. Use Zabbix HA with 2 nodes (point both server to DB on HA2)
2. Pause the active node (HA1) in a virtualized environment
3. Wait for the standby node to promote itself to active
4. Unpause HA1
Result:

Expected:
After recovery, HA1 should change to passive and not become active again automatically.