I can "behave well" and still be a bad actor in that I'm constantly spreading dangerous disinformation. That disinformation looks like signal by any metadata analysis.
Yes, that's probably the limit of the pure behavioral analysis, esp. if one is sincere. If they're insincere it will probably look like spam; but if somebody truly believes crazy theories and is casually pushing them (vs promoting them aggressively and exclusively), that's probably harder to spot.
