It should error out because apparently dings cause ert errors.  I did
not know that before but I am happy about it.

Eli Zaretskii <eliz@gnu.org> writes:

>
>> Also that situation doesn't respect completion-fail-discreetly if you
>> where to type in a failing completion before the final TAB press.
>
> Recipe, please.

I said this from a static analysis point of view but I couldn't seem to
reproduce this with a test.  It might be a dead code pathway but it
would still be nice for it to be correct.