If we assume he "spazzes" with 44, 77, A4ss, A7ss, and KTss and maybe some other pairs and suited As, how often would you think the average random 1/2 player is doing that from the BB against a player they basically know is going to be all in? Against a player they know it's going to be all in and a person that called in between? Now how about doing that and calling a 4bet back raise? Now how about doing all that AND playing to the river the exact way the hand went down? So you see the problem yet?
While you say the probability adds up, the multiplicative probability of all the things starts to diminish quickly. Let's say he can do this with one of those goofy hands 10% of the time, and I think that's generous for a random 1/2 player. And let's assume the flop mistake is not an angle, because it usually isn't. Then he decides to not raise turn with the sets only 30% of the time and it's basically always calling with the nut flush draw hands. For sets of 4s and 7s we are at like 3% of the time. And for nut nut flush draws we are at 7-10%. But the other 90% of his range are reasonable hands. It's not worth taking weird lines against these 10% kind of things. And I still think 10% is overestimating.
10% is high for a random player, i agree. but let's go through with details...
let's pretend 80% of players are pure nits and only 3bet AK and JJ+, and they always 5bet with AA. in fact they often don't even 3bet here with TT/JJ and AK, choosing to call half the time, and sometimes folding AKo and JJ to the 4bet. their initial 3bet range is about 2.5%.
- KK (6 combos) is not snap calling the river or giving that speech on the flop so we'll eliminate that.
- QQ (3 combos) queens will give this speech infrequently but could be as high as 1 in 6 times, ill say. call it
0.5
- AKss (1 combo diluted to 0.5 pre) is doing this most of the time on the flop and turn but not always since it may fold the flop, will once in a rare while bluff with a monster draw on the turn (barely impacts the math either way), and most importantly is much less likely to snap call on the river than the others. we'll call it
0.25
- AKs not spades (3 combos, diluted to 1.5 pre) likely to float flop, but still highly unlikely to call turn and probably won't snap river.
0.25
- AKo (12 combos diluted to 4 pre) makes no sense to call the turn, so we dilute that again and might not snap the river, so we'll call it
1.
- JJ (3 combos diluted to 1 pre) plays it like this every time on the flop, and about half the time on the turn. so we'll call it
0.5
- TT (3 diluted to 1.5) which never calls the turn or snaps river so zero
a total of about 2.5 weighted combinations, of which AKss i'd say is about 1 in 6.
so when you said "Suits are far from irrelevant. In fact, the whole point of the exercise is that you can actually almost narrow down the exact hand based on the suits in play."
... this isn't really true even if you're only looking at straight forward nits.
the broader point though - when you filter in the small number of spazz players,
there're about 20 combos of hands like nut flush draws (paired or otherwise), sets, QJs or other wonky two pairs that could reasonably play in this manner. only 1 in 5 players is ever 3betting these, and they're only doing it about half the time, and calling the 4bet about half the time they raise. that still leaves you with about 1 weighted combo. and they may not play the streets all like this, so again you're dwindling it down to maybe 0.75x (i think the nut draws, the wonky aces up and especially the paired nut draws play like this very often, the sets only about half the time).
but my point isn't that they're common, just that they comprise a non-trivial percent of the overall range when all these things are taken into account. if we're sticking with the 2.5 combos of nit hands, then these spazz hands are about a quarter of a random players range given these assumptions. or about 3 times the overall representation of AKss.