I suspect a lot of the speed comes from structure specific serialization (avoiding reflect). This can probably done with less unsafe code, and for most use cases that'd be a better trade-off.
Yeah. I have found that for message passing systems the tooling around multiple message types is more important in practice than optimizing for a single specific type. You can solve this with unions/tagged enums or some bespoke "message type metadata", but you HAVE to solve it some way. On the receiver end, you need to have a good story for routing different kinds of messages.