Thanks for the info, and for the years of use! Letting the user clear the
captcha in a browser and then handing the session to the crawler is a nice
solution for this hard problem.

One thing: the first is cookies: the loader used to drop every #HttpOnly_ line
from a --cookies-file, and that is where a login session usually lives. I
fixed it in master on 18 October, so it misses 3.50.5 and lands in the next
release.

I'm working on a very simplified version of a browser-human-assisted form and
captcha thing. This won't solve complex problems like in your case but it may
help basic ones.