Our dedicated server (Windows, Chernarus, community/modded) dies abruptly about once per day. The RPT truncates mid-line with no script error so a pure native crash.

Two minidumps from 2026-08-18 (20:45:43 and 22:05:38 local time) fault at the identical instruction: DayZServer_x64.exe+0x3C1D3D, exception 0xC0000005, WRITE at address 0x7 - a member-field write through a null/freed object pointer. Faulting instruction bytes: F0 44 0F C1 73 08 (lock xadd - an atomic refcount/counter update).

In both dumps the RPT shows the crash occurred during NETWORK: Replicate game state <player> roughly one second after an entity deletion - in one case an admin-tool deletion of a moving scripted entity, in the other a CE cleanup removal during infected despawn churn. A survey of 13 days of sessions shows our abrupt deaths overwhelmingly end mid mass-deletion activity (<cleanup> batches, dynamic-event despawn waves), while sessions ending normally never show deletion storms in their final output.

Suspected use-after-free: an entity deleted mid-frame while the replication layer still holds a reference.

Secondary, lower priority: two further dumps (2026-08-23 20:00:07, 2026-08-24 04:00:04, included) fault at DayZServer_x64.exe+0x618519, READ of heap addresses, during scheduled-restart shutdown/world unload - a teardown-path crash with no player impact.

Note on modding: the server runs many mods, but script VMEs cannot produce a native access violation inside the engine binary, and the identical faulting offset across independent crashes on different days points at one engine-side code path.

Not deterministically reproducible - frequency data instead:

1. Windows dedicated server, 1.29.163709, ~30,000 CE loot entities, 20-40 players connected.

2. An entity is deleted while actively replicating to nearby clients (observed triggers: admin-tool deletion of a moving scripted entity; CE cleanup batch during dynamic-event infected despawn).

3. Within ~1 second the server may crash during "Replicate game state" with the signature above.

Rate: ~12 abrupt deaths over 13 days on our server; both fully-dumped crashes match the same offset. Happy to run diagnostic builds or supply further dumps.

Deleting an entity that is mid-replication should be safely guarded; the server should never write through a freed entity pointer. Server process should not terminate.