On Wed, Apr 24, 2024, 20:24 Eli Zaretskii <eliz@gnu.org> wrote:
.

It isn't suddenly, it's because you switched Eglot to the new
track-changes method, right?  It worked fine before that, with the
same Quail, right?  Or am I missing something?

I haven't much to add to this discussion where I'm being Cc'ed except that I generally agree with Stefan's stance and fix and that I am fairly sure this has always a problem before Stefan's track-changes.el framework, only that we did not know about it because the raw usage eglot.el made of b-c-f and a-c-f didn't sanity-check anything. You just got subtly wrong server behaviour because it was being misinformed, and ended up reconnecting to fix these things.