We have in the past worked with 2 blind users to improve things (@MarcoZehe and @ndarilek ) see:
We received advice from @stevefaulkner an expert in the field and acted on it.
We will continue to improve Discourse based on feedback.
As it stands the biggest pain point, which is a devil to solve is Email parsing. At the moment our parser requires the response is “at the top” if it is inlined or anything non-traditional stuff gets rejected. That tripped @MarcoZehe last time when he tried to respond. I would like to improve on it and we are logging all email failures so we can extrapolate better logic.