PokerSnowie range strats that simply can't be correct

I finally sprung for the Pro version of PokerSnowie, and... I have thoughts...

There are some strategies suggested that can't possibly be viable. Not just like my human brain can't possibly comprehend what this super-machine is suggesting, but like it just straight up ain't mathin'.

For example, if BTN in a 100bb heads up game limps, the range it recommends for BB to raise 2 times the pot is over 75% of hands and then if BTN 3-bets, then BB comes back over the top with a 1/2 pot raise over 66% of the time. So BB is sticking 30bbs in with 50% of any two cards.

A lot of the examples I've found in the first 24 hours of my subscription are in these "non-standard" spots (ie: situations that maybe wouldn't occur at equilibrium) but far from uncommon--and something that could very easily spotted and exploited by a thinking player.

In every case, they are perhaps a viable exploit against the type of players who are likely to limp the BTN in a deep-stacked game, but if you were actually to implement this strategy into a match any thinking player would simply stop raising for value and will just induce an insane amount of action from us. I understand Snowie isn't a solver, so it's not necessarily devising unexploitable strategies (well, "understand" might be a stretch because I'm not a computer scientist or a mathematician), so if it's been trained on these spots largely against players with specific leaks, I guess it's just devising strategies that exploit those leaks rather than coming up with a sound strategy.

This makes it hard for me to interpret what I'm seeing in a lot of spots. I suppose observing strategies that only work against the players who are likely to take the line I'm facing is good practical strategy?

Any and all thoughts welcome, either about how reliable you find Snowie, best ways to use the software, corrections on my understandings of the relevant theory and computer science, you name it.