Q & A and comments can (and should) be dealt with on their merits independent of whether the poster is a crank. I don't see a problem with questions about squaring the circle, or 3-line proofs of major theorems and conjectures. Rating and tagging and filtering should do the rest. It is a distraction to constantly focus ad hoc attention on the "cranks" or "bad users" and their postings instead of moving the discussion technology forward.
[original longer answer follows. Above short answer moved up from comments.]
I think that this site should not classify posters or postings as "mathematical cranks" any more than it seeks to classify the intelligence, mathematical ability or psychological disability of all the other users.
Because answers have a separate existence from questions, and can be linked directly, "don't feed the trolls" (quoted in the other answer from USENET days, which had more interlacing of the troll and non-troll postings) is obsolete. A crank, crackpot or troll question can lead to non-crank answers that have a separate value, such as explaining a common mistake or why the standard theory is built in a particular way. For example, the recent 0.999...8 thread had both a massively downvoted question and highly upvoted answer. If crankish posters happen to manifest antisocial behavior in comment threads or questions this can be dealt with as its own phenomenon in the same manner as for other users, without making any crankery assessments.
Not understanding mathematics, or not having a conventional understanding, is not by itself a problem on a questions site. Some of the non-crank users seem to have psychological or social problems of their own, as is inevitable in a large population. That's life. It is OK. We don't make formal or informal policies seeking to limit their presence on the site and can appreciate their mathematical contributions for whatever value those may have.