It is much easier to play devils advocate then having opinion you advocate for. And it is overly easy to use devil advocate tactic to make it hard to keep original topic, to force people to go to lengthy tangents or otherwise prevent them to actually make the statement they were aiming for.

If I intend to discuss complex enough real world problem in good faith, especially one with unclear parameters, devils advocate basically makes it impossible.

I agree, but I would put it differently. It is a lot easier to advocate a conclusion with full control over the assumptions, than discovering the boundary assumptions where the opinion no longer holds.

For example, it is easy to argue for "HTTPS is more secure than HTTP". But what I am often interested in -- as a curious and skilled IT person, and not a lambda user -- is what attack models does HTTPS defend against. Remove assumptions one by one, when does HTTPS offer equivalent security to HTTP.

This is not just nitpicking, but extremely relevant for our profession. TLS has gotten under quite a lot of fire lately, so sooner or later, someone somewhere will discover those hidden assumption. I rather it happens on a public forum.