RTMPS Traffic

[Connect]

I am trying to help a network team (from an application perspective which makes me a knuckle-dragger I know) figure out why, since the introduction of their FortiGate (5.2.7 build 0718), their virtual Seminar hosting collaboration team cannot get Adobe Connect sessions (RTMPS) to go above a couple of hundred users without a surge of disconnects well into any seminar (maybe 60 minutes in on average). The Connect logs show the disconnections of participants without any server-side symptoms (such as fast-fails or PID changes to denote fail-over to new a backup server) to account for the drops. During a recent incident, 100 users were dropped from a 300 user seminar due to a network anomaly that we cannot identify. Is there a Fortigate firewall setting (perhaps application shaping or intrusion detection or keepalive option) that could trigger a timeout that would cut off a subset of seminar attendees running on RTMPS streaming sessions? Where on the Fortigate firewall might I suggest they look for a possible brilliant hidden feature that may be decapitating Connect at its RTMPS stream extended neck?

[MikePruett]

Does this occur when you make a policy specifically for RTMPS that doesn't have any layer 7 UTM features on it?