update: shamir secret sharing is post quantum that's good news, and I now suspect that 3 (weak password) probably isn't a problem.

Hmm, the paper I read how-not-to-share-a-secret.pdf is actually about a different scheme to shamir's, that seems to allow sharing multiple sets of secrets at once.

reading Detection and identification of cheaters in (t,n) secret sharing scheme detection-of-cheaters-in-sss.pdf

outlines 3 attacks/problems

1. a peer returns an incorrect share by accident (for example, because of a bug)
2. malicious peers generating incorrect shares, but without having the chance to see any honest shares first.
3. as in 2, except attackers have the chance to see what honest peers reveal first.

3 is obviously a lower bar (thus easier to attack than 2) and is also more relevant to dark-crystal.

Hmm, okay I don't fully understand this paper (too many cryptic math symbols) but they claim they detect cheating in O(1) and detect who the cheat is in O(j!) (I think j is the number of cheaters). j factorial is pretty big, but okay since j is small. It's the same as my suggestion would be though. (I need to look at this more though)