Even if we ignore any ethical concerns, when you’re sampling from an extreme environment, the strains you’re finding will, with >99.9% certainty, have substantially diverged from a biologically identical ancestor that’s spent a fair number of generations infecting hosts.

So you also get a weird Ship-of-Theseus type question of “are these still really the same bacteria?”. And if you assume they are going to be different strains after adapting to different environments, then you can also safely assume that whatever strain you’re sampling in an extreme environment has a >99.9% probability of being in the potentially-harmful, contextually-harmful, or non-harmful bucket, by virtue of the fact you found it isolated in the wild rather than in living hosts.

To put it a little more simply: if you’re looking for something with a demonstrated ability to infect people, you’ll probably find that inside or nearby people, not in an icy, remote cave.