Erlang style "fail noisily and let the supervisor handle it" is quite a solid way to deal with "exception handling bloat" in any paradigm, functional or otherwise... The hard part is finding a way to create effective layers of supervision.
An exception is a noisy failure, and the supervisor is whatever try/catch that it bubles up to. Still, Erlang style has it's own complexities. It is in essence a network of micro services, and creating that overlay network and manuvering in it isn't simple.
Yep.. with python assert statements + supervisorctl has been a very good setup for scaling a low latency, clean chat filtering app.