There’s likely some cases of convergent evolution, but I’m not sure this is settled for all trees.

I’d suspect that at least some trees with last common ancestors that are shrubs have re-enabled genes that enable the tree phenotype rather than independently evolving the tree phenotype.

But still really cool (and maybe turns on exactly how much evolution your consider needs to happen before it’s convergent evolution.)

Exactly.

The general approach is to use interpretable models where you can understand how the model works and what features it uses to discriminate, but that doesn’t work for all ML approaches (and even when it does our understanding is incomplete.)

Maybe not the hardest, but still challenging. Unknown biases in training data are a challenge in any experimental design. Opaque ML frequently makes them more challenging to discover.

There’s even an instructables on how to do it.