Theoretically, which is why it was recommended to prescribe as a last choice for those who couldn't use, or didn't respond to, other treatments. - see mikeyouse's comment. (Though that doesn't seem to be the only way it was prescribed.)

The authors of the study haven't yet identified any molnupiravir-associated lineages that spread to more than 21 people.

Anything at all which inhibits the virus's spread will do this. That is extremely basic evolutionary theory.

Speaking of theories, I have a theory that most people don't actually believe that evolution is real, because they question basic results like this. The only reason they profess to believe that evolution is real is because that is the only socially acceptable option.

Depends on the population, it means that they increase the number of drug resistant mutants. Since it only affects people who would take molnupiravir which are then exposed to the mutant anyway (likely in hospitals and rest homes?) the alternative is no treatment. However, in these patients the no treatment alternative is likely to lead to much longer infections with more mutations and chances for transmission.

If those people were well isolated when they were taking the drug then the mutants wouldn't be transmitted. If the hospitals and homes are well isolated then the mutants won't be transmitted. There is in any case a wide reservoir of virus and mutants in the broader population and it isn't clear that in that broader population that this resistance makes the virus fitter (more likely to be transmitted) and it's likely they are less fit.

The really infective versions seem to have come from immune compromised patients who were infected and likely infectious (not just sick like long covid) for months.

Yes. And the linked study shows that this is not merely theoretical, it has in fact already happened.