We started experiencing significant performance degredation around 4 days ago. Other users commenting on this thread are reporting a similar degradation starting around that time.
We experienced this across several games, including some which had not received updates.
We have also increased the server capacity from 150 to 160 players on November 7.
We had no new feature changes that were a direct result of introducing new gameplay scripts or modifying existing ones in a way that required adding more connections per player/tracked instance.
On November 9, we identified a bug with most of our game scripts failing to function as intended and have shipped that immediately.
We started noticing performance issues on November 11. I was paged to a memory leak issue on 11/11/2025 at 1700 UTC, so the performance issue could’ve started much earlier.
The update pushed at around 0000 UTC on November 12 addressed some memory issues that we think may have caused the problem.
We have noticed some performance improvement, but only within the first 6-12 hours of the hotfix publishing. We have attempted a fix on November 13, however that didn’t seem to address the issue any better or worse than November 12.
I should note that games that we are noticing a degraded performance do use parallel LuaU (actors) for server-sided hit validation and AoE (area of explosion) hit registration, which are both computationally heavy.
Apologies for the long silence on this thread (this post was missed somehow), and thank you for the detailed write-ups and microprofiler captures. They were genuinely helpful in tracking this down.
You were likely right that this was tied to the server CPU allocation (“More Server CPU” / dynamic CPU) rollout. In the weeks after your reports, we identified and fixed several distinct issues that had surfaced around that same window:
A datacenter-level problem where there are periodic network/firewall issues. This was resolved shortly after it appeared.
A bug in the new dynamic CPU system, where a server’s actual CPU quota could stop updating as players joined and left. This lines up with the “performance gets worse as the server fills up” behavior several of you described.
The final fix for the CPU-quota bug was fully deployed to production in early 2026. It self-corrects for servers already running, so there’s nothing you need to change on your end. Since then, multiple affected developers with high-population experiences have confirmed their performance returned to normal (including running 200+ player battles on a single server again).
Given that, I’m marking this report as resolved. If you’re still seeing degraded performance on current servers, please reply here or open a new report with your place ID and a recent microprofiler capture that would most likely point to a separate issue, and we’ll take a look.
Thanks again for your patience and for the thorough reports.
