We are experiencing a fatal memory crash that forces the zabbix_server service into a continuous restart loop.

Steps to reproduce:

1. We disabled "Use combined requests" (bulk polling) on a heavily polled SNMP device, which flooded the engine with individual requests.

1. Simultaneously, multiple switches on our network are improperly sharing the same SNMPv3 Engine ID.

1. The libnetsnmp cache attempts to process the duplicate Engine IDs under high concurrency.

1. The Zabbix server memory cache corrupts and triggers a Segmentation Fault.

Logs: Prior to the crash, the log rapidly prints: SNMP engineId is not unique across following interfaces: <IP_Address_1> and <IP_Address_2> Immediately followed by the server crashing with a SIGSEGV error.

Workaround Applied: Setting StartSNMPPollers=0 stabilizes the server by bypassing the SNMP engine entirely. Running zabbix_server -R snmp_cache_reload and disabling the duplicate hosts allows us to restart the pollers safely. We are requesting further stability patches for the SNMP caching engine to prevent duplicate hardware IDs from fatally crashing the entire monitoring service.