Many thanks for your comments.  I attach the updated patch, which
contains a few other corrections and hopefully not too many regressions.

> AFAICT it should be safe in the sense that it should not affect behavior
> in file buffers.  See comments below.

That's indeed the plan.

> Should we push it to `master`?

I'm curious what Arash/Ikumi think -- if they don't like the idea, then
I'm happy to put this in the wastebin and regard it as an exercise.

Also, if we're going to proceed with this, then I'd like to take the
opportunity to add support for indirect buffers to RefTeX, which
requires modifications to the same parts of the code as this patch.

Thanks, best,

Paul