Interpretations can't be wrong, otherwise they would not be interpretations. They also can't be right, for the same reason.

Two questions. First, is that true as a matter of how you define 'interpretation', or is it true as a matter of subsequent fact? Second, do you mean to say that interpretations haven't yet passed or failed an experimental test?; or do you mean to say that interpretations can never pass or fail an experimental test?

The 'haven't yet' criterion is weakly true of 'interpretations' in the relevant QM context, though all the interpretations of interest here have been verified relative to empirically false models; they just haven't been verified relative to one another. But the 'can never' criterion is clearly false of some of the 'interpretations' we're talking about, and only contingently and ambiguously true of any of them. Whether these models will ever be empirically testable is itself an empirical question.

Here I define"wrong" in the natural sciences sense, "failed an experimental test". And that's the only definition that matters when talking about QM (which is a natural science)

What are you building into 'right'/'wrong' here? That is, why does your assertion have more semantic content than if you'd skipped the 'right'/'wrong' assertion and just said 'The Bohm interpretation hasn't passed an experimental test yet. Bye now!'? Certainly if you mean to suggest that scientific models cannot have merits or dismerits aside from experimental verification/falsification, then this is wrong. Scientific models can be overly vague, or ambiguous, or internally inconsistent, or overly complex or inelegant, or unexplanatory, or gerrymandered, or ad-hoc, or unverifiable, or unfalsifiable, or historically (as opposed to experimentally) false. All of those are faults in their own right -- and, often, they are Bayesianly relevant faults, faults that should impact our credence in the model.