Basically that, yes. The idea is, as long as the mutable state starts and ends within the confines of an individual function, you can get away with treating that function as if it were "truly" FP.
The main argument against this is that it's a concession that makes the codebase less coherent. The more hardline argument being that the encapsulated mutable state is still more prone to bugs and should be almost always be avoided, except at great cost.
There are certain algorithms that either can’t or only insanely error prone way can be expressed “purely”. There is nothing wrong with locally mutable state. It is basically the same as a State monad that is evaluated/executed — which can arguably be worse
It’s rare that you need to use vars in Scala, but still if it’s scoped to the function then you know you can mentally discount it once the function terminates.
The main argument against this is that it's a concession that makes the codebase less coherent. The more hardline argument being that the encapsulated mutable state is still more prone to bugs and should be almost always be avoided, except at great cost.