Hi All:
The issue of the moment is gnome-speech, which I believe is still talking OSS in order to also support non Linux architectures like Solaris and Aix, which do not have ALSA.
The issue is actually the speech engine itself. gnome-speech itself does virtually nothing with respect to audio -- it merely is a wrapper on top of the underlying speech engine. As such, it is the speech engine provider that needs to provide better support for audio device contention/sharing.
Will