Stress testing, I created some 40,000 parts on the server in a Heartbeat connection on the TaskScheduler. Heartbeat does not have a queue, so in Studio allowed execution time gets eaten up for the instant and the (Luau?) stack gets exited. If not in Studio, AFAIK, allowed execution time doesn’t exist, so the server just crashes.
What’s really interesting is on the client, you see the parts spam out over the course of several seconds before crashing, with an average replication level of 360kb/s . What determines this throttling queue rate? Is it arbitrary or systemic?
1 Like