In regard to bots. Could we identify a gto strategy if could see all of a certain players hole cards? Because I think if we could then we would have solved poker. It seems it would take a bot to know a bot by identifying the spots the suspect played a certain degree of differently from how the detective bot would (perhaps that's not even good enough). Regardless, the cheater could dial down their bot so that it is not so profitable that it gets caught. Bots seem not very detectable.
Moreover, I think we can compare bots to bitcoin miners, as they are trading computational effort for transferable utility (money if we consider bitcoin and crypto-currencies money). And in this sense the old mantra truly applies, they are generating rake (regardless of whether the operator prefers to have bots or not).
Consider a randomly seated pool of bots playing in a wild wild west field. I think these bots wouldn't grind to the point of the games being -ev for them (assuming they all basically have the same winrates). This was actually one project's solution (Pokereum) which used random seating as part of their cryptographic protocol security algorithm.
Thinking of this as the base layer, then it should be, since its randomly seated, decentralized in a certain fashion. And this would allow for a proper jury pool for hand validation (obviously you would have bot pools arise).
The extra step on top of Pokereum is that this layer of bots secures the cryptography for the layer above it-private sites.
It COULD be done such that the private sites subsidize the underlying bot layer which would still be an obviously incredible net savings to the industry.
I basically conceived this before ethereum was implemented (probably not before I read of it).
The resolving consideration is the equilibrium between the profits bots could make mining either ethereum (or other crypto-coins) or playing/mining poker.