Hi!

It seems that 25 fixes the problem. The bug does NOT occur with
Emacs-2016-01-24_01-41-01-8cdb9ac0fb060366bc823de84ecdba258716addb-universal.dmg
from that address. I also checked the build of 24 up there (
Emacs-24.5-1-universal.dmg ) and it still has the same bug, but I can
confirm that the bug is gone in that nightly build of 25.

This is great news! I will close the bug, but I will investigate the cause of the regression I noticed.

I'm totally happy to help debug the problem in 24, as it would be an
excuse to understand the development process better, but I'm also happy
to wait for 25.1 to become stable, which seems imminent.

Say the word and I'll figure out how to compile a debugging version of
24 (I just managed to codesign gdb), but I suspect that energy may be
better spent on 25.

No need, as it is fixed in Emacs 25.

Of course, if you feel like contributing to NS interface in Emacs 25, you're more than welcome, as the long time maintainer retired last year.

-- Anders