Should I ask a new question when while testing a recommendation it doesn't fit some needs I discovered later on?

Question

While testing some recommendations I noticed some problems intrinsically of the requirements of my question. I might need to add requirements, rewrite others or just remove them. The answers do meet my original requirements completely but I found out other problem unrelated to the tools which I asked for, that will make them useless for my proposes. Should I ask a new question or rewrite and possibly invalidate the answers of my original question?

Answer 1

An alternative approach to the kind of "update" I proposed in my other answer would be creating a new question. But to avoid "VTC as duplicate", that new question must include a fitting hint, like

This is similar to {question A linked}, but has additional/slightly-different requirements ...

This approach would avoid effects on already existing answers completely, as the context of the original question remains untouched. After all, it's not the answerer's fault if you missed pointing out all your requirements (but neither yours, if you only got aware of them when trying offered solutions), so it might be the better approach. After all, I don't see a problem with related questions, as long as they aren't "identical".

Answer 2

That's a tough one. If there are no answers yet, it's easy: update your question accordingly. But if there are already answers which your update would "invalidate", that might prove unfair to the answerers: Later visitors might not notice you've changed your question when those answers already have been made, and downvote them for "not matching the requirements".

One possible approach would be adding new text at the end of your question, starting with a "horizontal ruler" and a bold UPDATE (date/time):. A good idea here might be even to mention something like "Note: answers X, Y, Z have already been made when I wrote this update", to give them a fair chance.

Of course, there are always people who still "wouldn't get it", so this might not be the best approach.