Generally, question edits that invalidate existing answers should be rolled back. There is plenty of precedent for this across the network.
I haven't looked at this particular one in detail, so I can't really say if it qualifies for being rolled back. But from even a very brief look, it certainly looks like a solid candidate...
If a question has answers and an edit not only invalidates those answers, but also completely replaces the question (the sheer size of its revision 3 diff is telling), then roll back that edit and tell the poster to ask the different question as a new question.
That way, people who spent time answering the old question don't look like they're answering something completely different than what the OP is asking, and both questions can get the attention they deserve.
Sometimes, we (yes; myself, too) find out only after asking a question that we actually asked a question that was different from what we intended to ask. That doesn't mean that the question we did ask does not have value, nor that the people who did answer that question should have their effort taken away from them. The best course of action then is almost always to ask a new question, and in doing so take extra care to actually ask the question that we want answered.
As for the question in your title right here, only actual duplicates can ever reasonably be candidates for merging, and not all duplicates will be candidates for merging. If you can no longer vote to close (because you have, say, already used your vote on the question), then bringing it up on Meta or in chat is the right thing to do, but for review and possible closure as duplicate, not for merging. But in this case, I believe the first step before doing that would be to roll back to the original revision (for the reasons and as discussed above).