[Tts-project] Bug#859926: breaks with pulse-audio as output when spawned by speechd-up from init system

[Cobra]

Control: found -1 0.8.6-4
thanks

Hi,

did another test:
The installation and start/stop/restart of speechd-up work when
changing /etc/speech-dispatcher/speechd.conf from

# AudioOutputMethod "pulse"
to
AudioOutputMethod "pulse,alsa"

But doing so will break pulseaudio for users. There will be no audio
output from applications like vlc or firefox until speechd-up is
stopped. vlc is showing only "dummy output" instead of the usual
"Built-in Audio Analog Stereo" in Audio/Audio Device when speechd-up is
running.
Starting speechd-up after stopping it and using pulse as a user fails
until the running pulseaudio daemon is killed.
Switching vlc to use only alsa (both output module and device) works
with speechd-up running.

This whole issue looks very similar to #521675 (pulseaudio), note the
mention of speech-dispatcher which turned out to be the cause for one
user.
There are also #625235 and #670740 in speech-dispatcher.
A solution might be running pulseaudio as system-wide daemon. But I
don't know whether this is desirable or not, because there are some
downsides, including security implications.