I think an example about a failed dead startup is most informative about startups in general.

What's that? Surely if a prophecy were a useful filtering mechanism as you say, then dying is a problem. A cult which fails cannot serve anyone's purpose at all...

I'd be better served by picking a reputable book about cults at random, anyway (cherry picking vs unfiltered data).

Fair enough, but shouldn't you then retract your previous claims? I mean, what with it being based on cherry picked evidence and all?

Furthermore, sunk cost fallacy - like failure mode seems massively relevant to retention in cults.

You should probably know that I consider it seriously questionable whether sunk costs affect individuals at all, then, and reject the premise of that argument, much less whether it applies to cults.