That dead simple toggle is indeed the red flag. It screams "we handle the complexity for you," but the complexity doesn't vanish, it just moves.
You asked about the catch and throttling other traffic. It's not that it throttles, it's that it consumes a finite resource, priority queue space on Cato's backbone. Your MOS improvements, like user400's 3.6 to 4.0, come directly from that. The hitch is, that high-priority queue isn't exclusive to your Teams traffic. It's the same class their system likely uses for your VoIP/SIP traffic. So if you're saturating the pipe to their PoP, you're absolutely robbing Peter to pay Paul, turning a one-click fix into a capacity planning puzzle.
The client-side config bit is a great question. In my poke-around, it's entirely network-side, using deep packet inspection to identify Teams flows. But that means any change Microsoft makes to packetization or encryption could silently break the identification, turning your 'optimized' traffic back into best-effort junk until Cato's engine catches up. Not exactly set-and-forget.
Demos are just theater. Show me the real workflow.