I think doing a few 100 is good. It teaches you some patterns and ideas.
But there are a fairly limited set of patterns / ideas. I think solving from a book is better.
And when doing timed, it's too easy to guess moves because there are never any subtle or difficult moves. It's (more or less) always a check or capture... so you can just make a random sacrifice or capture and hope it works out.
I've seen some low rated puzzles like this. Where the engine defense is really stupid. For example you sacrifice a rook and it just walks into a mate in 2... but in a real game to sacrifice a rook you'd have to prepare for much more difficult variations. Sometimes low rated puzzles involve 2000+ justification (and I mean a 2000 live rating, not a 2000 puzzle rating) in the unplayed defensive lines... but in puzzle rush you'd be stupid to calculate the difficult defense because it would take too much time.
So yeah, the patterns are limited and it encourages sloppiness, but doing a few 100 is useful IMO.
I have been listening to chessdojo talks and they seem to think that puzzlerush is not good for chess improvement. I think its a great idea for short term improvement and pattern recognition for beginners and intermediate players, especially survival. Also survival improves stamina because although I would solve 2500 + puzzles with more of a success rate in rated puzzles I start getting tired in survival and stop calculating fully and survival negatively reinforces such behaviour with a cross