Issue
A two-player Studio test looks correct, but larger servers show corrections, CPU pressure, or unstable behavior.
Cause
The test does not represent real player counts, latency variety, physics ownership, or ability volume.
Fix
Use Roblox scale testing where available, instrument simulation cost and misprediction rate, and add players gradually. Test the worst combat or vehicle density instead of an empty map. Keep a rollback plan before enabling the feature for all production servers.
Verify
Compare server frame time and correction metrics at target concurrency against the current production system.
Official Roblox scale-testing release recap
#1
Original Heliotrope guidance with authoritative Roblox sources.