I've never used esd or arts since I switched my desktop to Debian about a year ago. I've always used just plain ALSA, and it's always worked great. For GNOME, you can use GConf to set the outputsink to use the alsasink, and in KDE you can simply select it in kcontrol. I really don't know what Firefox uses, or how it decides, but I've never had to change anything on it.