Looks fine, but would need, I think, a decent amount of testing:

* In eglot.el, being validated by both Andrii and you for your particular use cases.

* In bare emacs-lisp-mode, with multiple eldoc-documentation-strategy options.  It should

  behave exactly the same before and after the patch, as elisp doesn't use invisible stuff.

* In other situations you can think of.  Like variations on eldoc-area-use-multiline-p and

  eldoc-echo-area-display-truncation-messages.  Again, here the behaviour should be

  equivalent pre- and post- patch.

Just noting this because in my memory this code is somewhat hairy.  That

said, the patch "looks" really fine :-)

Thanks,

João