Thanks for the update. However, it seems you still want to go ahead with automatic reopening of questions and just plan to make it a bit smarter e.g. with regard to what close reasons are eligible. Let me just state my opinion on this:

###Questions should not be automatically reopened under any circumstances

I can guarantee you that whatever heuristics you use to determine if an edit was sufficient to re-open a question will simply not be good enough, at least not for my standards. Imagine the typical "plz halp debug" code dump question that gets closed because it needs an MCVE - probably one of the most common cases of closed questions that's not a dupe or straight up off topic. So what if the user edits the code, or the description? How would any automated system even begin to measure if the code is enough (and not way too much) to reproduce the described problem, or if the description is clear enough to warrant re-opening? To do so, for starters you would need to implement very sophisticated code analysis for all programming languages supported on SO, nevermind the NLP requirements to understand the problem description.

What you should do instead, is simply throw the questions which your automated system flagged as potentially reopenable into the reopen queue and let the users handle the task of actually judging if the question is good enough.

Thanks for the update. However, it seems you still want to go ahead with automatic reopening of questions and just plan to make it a bit smarter e.g. with regard to what close reasons are eligible. Let me just state my opinion on this:

Questions should not be automatically reopened under any circumstances

I can guarantee you that whatever heuristics you use to determine if an edit was sufficient to re-open a question will simply not be good enough, at least not for my standards. Imagine the typical "plz halp debug" code dump question that gets closed because it needs an MCVE - probably one of the most common cases of closed questions that's not a dupe or straight up off topic. So what if the user edits the code, or the description? How would any automated system even begin to measure if the code is enough (and not way too much) to reproduce the described problem, or if the description is clear enough to warrant re-opening? To do so, for starters you would need to implement very sophisticated code analysis for all programming languages supported on SO, nevermind the NLP requirements to understand the problem description.

What you should do instead, is simply throw the questions which your automated system flagged as potentially reopenable into the reopen queue and let the users handle the task of actually judging if the question is good enough.