Comment # 14 on bug 916048 from Troels Just

(In reply to Takashi Iwai from comment #13)
> The feature should be already there, IIRC.  However, whether it actually
> works depends on the hardware; some Intel onboard graphics have a fragile
> HDMI/DP audio notification (it's rather a hardware bug), and might not work
> always reliably.
> In the kernel side, the audio device for HDMI/DP is always present no matter
> whether it's plugged or not.  If pavucontrol shows the HDMI device (plugged)
> or (unplugged) properly, the switching should work as is.

pavucontrol does indeed indicate the status of the HDMI audio connection,
however no automatic switching takes place at all.

I checked in /sys/class/drm/card0/card0HDMI-A-1/status, and it consistently,
and reliable indicates whether HDMI is plugged in or not. This is the case both
on my T420 and my X230.

I looked around a bit, and I found some tricks where a udev rule invokes a
script that then switches PulseAudio's output depending on the status of HDMI
connectivity in /sys. I tried it, and it does seem to work with one problem.
udev does not invoke the script when I plug the cable in (Despite /sys
indicating the status correctly), so I invoke the script manually and it works.
When I disconnect the cable, udev notices without a problem, script is invoked,
and sound is returned to the laptop speakers.

Here's a thought, could it be that udev not noticing the plugging in of HDMI,
and PulseAudio, on its own, not doing any switching are somehow connected?