Steps to reproduce:

1. Set up Zabbix HA servers, 3-way, with psql (16) and timescaledb (2.15.3)
2. Offload collection of data to proxy groups, and several proxies hosted in various k8s clusters
3. Use loads of IPMI and SNMP items (~3000 vps)
4. Let the server be primary for an unknown amount of time

Result:

We didn't manage to get a screenshot but, the Zabbix server HA node was marked as `Unavailable` (red text), in the "System information" widget, with "High availability nodes" selected.
 The zabbix_server process (and server port) were up, but they were not accepting/persisting data to the DB.

Log from the faulty node attached
Memory dump grepped from the log, attached separately

Expected:

While a crash is obviously no desirable, we would expect the entire Zabbix server process to properly crash instead of ending up in a state, where we don't accept data, and don't fail over another server node.