Villain is a regular I've only played with once before. In this session, he's in the process of burning through at least $1,200, roughly $200 at a time. At the time of this hand, he's probably three buy-ins deep.

We have one relevant hand in our history, where he opened to 12 in EP and I 3-bet to 40 next to act with AA. He called. Flop T-9-x. He checked. I bet 60. He called. Turn T. Check, check. River x. He bluffs 60 with 9-crappy kicker, and I call.

This guy is super active. He plays almost every hand, often coming in for a small raise preflop. He will seemingly always call a 3-bet, regardless of position. The way he was playing, it's actually amazing that he didn't burn through more money faster.

Villain has ~270 (hold that thought); I cover.

Button straddle for 6.
One of the blinds calls, Villain calls in EP, I raise to 30 next to act with KK, Villain calls.
Flop (73): K94
[Villain checks dark, I bet 45 dark]
Villain check-raises all in, I call.

*Note: The dark bet thing is something I'll do maybe once or twice in a session. I do it mostly to mess with people and perhaps make it seem like I'm much crazier than I actually am. In retrospect, I don't love it in a straddled pot, because it's much more volatile.

I called his shove after just eyeballing his chips rather than getting an exact count, which was probably kind of stupid. I thought he had less than 200. In any case, my question is this:

In this scenario, how much must Villain have in his stack before we consider folding?

In the worst-case scenario where he flopped a flush, I'm 35% to win the hand.

So in the actual hand it was $195 back to you when he shoved? You're getting roughly 1.8:1 on a call? Since 2:1 is an automatic call then your spot seems like an easy call given that his range is wider than just made flushes. I would say any more than $250, putting it at roughly 1.6:1 to call and it's a fold vs. most villains, adjusting slightly for villain spew factor accordingly.

If villain is really active but won't bluff-shove with air on a monotone flop, then probably never once there is a substantial pot. There are just so many ways he can have a flush draw, smaller set or two pair. If your chance of drawing out was 1/3, and villain shoved for the size of the pot and showed you a better hand it would be 0EV to call. With a 35% chance to suck out your break even against over pot shoves. In fact, your break even point is right around the $195 villain has left, it's just about 0EV to call here even if villain shows you a flush.

I can construct situations where I could fold a set to one villain on a monotone flop. They require nitty/passive villains, deep stacks and small pots, and even then I'm probably not folding top set.

The thing is, if the villain can gii in here with sets and nut flushes, then it's very unlikely that we can fold for most reasonable stack sizes here.
If we think that he can do this with AXx, KxX type hands, then we can never fold.
We are 75% against AX, 90% against a lower set, and 98% against any two pair. So as long as we have 1 combo of anything else in the in their range for every 4 flush combos in their range, we can't fold.

I'm def snap calling here given the described action.
If V is offereing 1.3:1 or better, I'm likely not folding in this situation. After that, I think I would need reads to augment my decision.

Looks like it's pretty close if I can somehow put him on nothing but flushes. I'm approximately a 1.85:1 dog against flushes and the pot is laying 1.84:1. Of course, it's very difficult to put him on exactly a flush, especially given the way he has been spewing money into pot after pot.

Result: I called, and he did in fact have a flush. 52. The board did not pair.

To the surprise of no one, he proceeded to set fire to those winnings and then some.