I believe that the use of homographs definitely suggests the intention to hide plagiarism and so a form of ASCII binding could be set up to deny the use of homographs. This would prevent plagiarized character strings appearing identical to the original if a web scraper or bot was to view the strings, but the scraper should still be able to tell that 99% of the text is copied.

I'm not sure if Bitcointalk has a bot that checks for plagiarism but I assume it does, (or else the use of homographs here wouldn't make a difference) and so another alternative would be to check any text that uses 'mimic characters' with more scrutiny. For eg, a string without the use of homographs could clear a plagiarism check if 70% of the text is original, but a text with homographs might only pass if 90% is original, or not pass at all because the only logical reason for the use f homographs is to evade recognition.

A third option would be to just report simular looking posts to moderators, but even here, the use of homographs to conceal plagiarism is negligible.

So the use of homographs logically means that it is the posters intention of hiding plagiarism from bots so the best course of action would probably be to implement some function into the web crawler that check Bitcointalk and give it permission to delete and ban all posts and accounts that use homographs. The Armenian characters set has characters identical to the Latin character set, eg. o, n, u, S and Լ. There isn't a Latin board here, so removing Latin characters from this forum could work.