Interesting approach here.
I doubt OCamlers will want to drown in monads, monad transformers and inefficient data structure, though. There seems to be a stronger movement in the direction of Eff.
Puritas, A journey of a thousand miles towards side-effect free code (just the 29 first slides)
As in many presentations on this topic by people pushing purity, this presentation conflates I/O side effects, which are truly ‘dangerous’, with state-based side effects, which are just a source of additional complexity (and efficiency). I/O side effects can be dealt with by other means, making the argument mostly moot.