Summary
FiveM client (game process) reliably hangs and crashes after ~18-20 minutes of active, connected gameplay. The hang is caused by an access violation (invalid pointer write) inside citizen-devtools.dll, in fx::ResourceMonitor::SetShouldGetMemory. Reproduced across 10+ separate play sessions, on two different servers, with essentially every environmental variable eliminated.
Crash signature
FAILURE_BUCKET_ID: INVALID_POINTER_WRITE_c0000005_citizen-devtools.dll!Unknown
SYMBOL_NAME: citizen_devtools+2ccd6
MODULE_NAME: citizen_devtools
IMAGE_NAME: citizen-devtools.dll
IMAGE_VERSION: 1.0.0.32556
ExceptionCode: c0000005 (Access violation - Write)
ExceptionAddress: citizen_devtools!fx::ResourceMonitor::SetShouldGetMemory+0x12776
Attempt to write to address 0x0000000000000DEED
citizen_devtools!fx::ResourceMonitor::SetShouldGetMemory+0x12776:
mov dword ptr [0DEEDh], 0
crashometry.json on every capture also reports "crash_hash": "citizen-devtools.dll+2CCD6".
Observed lead-up (from CitizenFX log, every occurrence)
Starting ~18-20 minutes after entering a hosted session, the following repeats every ~1.5s until the watchdog kills the process:
Error: SendNetPacket -1276497698 called from multiple threads!
... (repeats every ~1500ms)
Window Watchdog/ Error: FiveM has stopped responding
DumpServer/ Process crash captured.
The specific packet-hash number in the SendNetPacket error varied depending on which server/resources were connected, but the pattern (fixed-interval spam of the same hash, starting ~18-20 min in, followed by watchdog kill) was identical in every reproduction. This looks like the ResourceMonitor’s periodic memory-sampling tick eventually dereferencing a stale/invalid pointer, taking the main thread down with it.
Reproduction
100% reproducible: crashed in every single test session that ran past ~18-20 minutes of active connected gameplay (10+ attempts). Does NOT happen while idling in the FiveM main menu (not connected) - only while connected to a server with resources actively ticking.
Environment
- FiveM client, GTA5 build 3258 (also reproduced after switching client UpdateChannel to canary, and after forcing a different GTA5 content build via sv_enforceGameBuild - see elimination list below)
- OS: Windows 11, build 10.0.26100.1
- OneSync: tested both enabled and disabled - no difference
Hardware
- Laptop: ASUS ROG Strix G16 (G615LW)
- CPU: Intel Core Ultra 9 275HX (24 cores / 24 threads)
- RAM: 32GB
- Hybrid/switchable graphics: NVIDIA GeForce RTX 5080 Laptop GPU (driver 32.0.16.1074, 2026-07-02) + Intel integrated graphics
- Active ASUS power profile at time of crashes: Silent (aggressively limits CPU/GPU power draw)
Other players (on desktop PCs, single discrete GPU, no hybrid-graphics switching) connecting to the same servers do NOT experience this crash. A laptop-specific factor - most likely GPU switching/Optimus-style hybrid graphics interacting with the resource monitor, or power/thermal throttling under the “Silent” profile - looks like a plausible contributing condition on this machine specifically, though not confirmed as the root cause: see the related report below, filed by a desktop user (Ryzen 5 3600X / RTX 3070) with the identical failure bucket, so hybrid graphics is unlikely to be a strict requirement for hitting this bug - more likely something that changes how quickly/reliably it’s hit.
Related existing report
FiveM randomly freezes | INVALID POINTER WRITE c0000005 citizen-devtools.dll (Mar 2024, desktop PC, Ryzen 5 3600X / RTX 3070, 0 replies) reports the exact same FAILURE_BUCKET_ID, but with random timing (5 min to 1 hour) rather than the consistent ~18-20 min window seen here. Together these suggest a real, long-standing engine bug rather than something environment-specific - our testing here additionally rules out basically every resource/software/config variable, which that report didn’t get to.
What was ruled out (each tested independently, crash reproduced identically every time)
- Four separate third-party vehicle resources with obfuscated/incomplete companion scripts (car2, plus three others referencing missing loader dependencies) - removed entirely from the server, crash persisted with the exact same SendNetPacket hash.
- Testing on a completely different, unrelated public server - crashed there too.
- VPN/system-wide proxy tunnel (Mihomo/Clash-based) - fully closed, crash persisted.
- Background OEM software (ASUS Armoury Crate / GameFirst / Optimization services) - stopped, crash persisted.
- Full FiveM client cache wipe (FiveM.app/data/cache + server-cache-priv, ~1.8GB) forcing a complete re-download of all assets - crash persisted.
- ASI plugins / injectors: ReShade 5.x (flagged by FiveM’s own CitizenFX.ini as “has a bug that will lead to game crashes”) and mod-menu ASI plugins (Menyoo, LambdaMenu, etc.) - entirely removed from FiveM.app/plugins, crash persisted with byte-identical citizen-devtools.dll+2ccd6 failure bucket.
- Client UpdateChannel: switched production to canary - crash persisted with the identical failure bucket after a full platform update.
- GTA5 content build number: switched from 3258 to 2372 (via sv_enforceGameBuild) - crash persisted; note the actual game session still reported running under the exe 3258 platform components regardless of the enforced content build, suggesting citizen-devtools.dll versioning is decoupled from the GTA5 content build number.
- Server-side resource count/composition: reproduced both on a heavily modded server (~780 resources) and on an unrelated third-party roleplay server with a different resource set.
Files available on request
Multiple full .dmp minidumps and matching CitizenFX_log_*.log files (each showing the identical lead-up and failure signature), plus crashometry.json for each crash. Happy to attach a zipped sample set (dump + log + crashometry from 3 separate reproductions) on request.