Bummer!

The comma got me on the right trail afterall.

tazhw detects the required modules with > lspci -k | grep "modules" | cut -d ":" -f 2 | sed s/-/_/g < which unfortunatelly reports p54pci and prism54 as a possible module for this card.

 snd-maestro3
 yenta_socket
 yenta_socket
 p54pci, prism54
 
Indeed both modules report themself in /lib/.../modules.alias as the "choosen one" for this card. :-(
alias pci:v00001260d00003890sv*sd*bc*sc*i* prism54
alias pci:v00001260d00003890sv*sd*bc*sc*i* p54pci

Cutting off the 'additional' options by adding '| sed s/,.*//' to the detection line in tazhw will fix the problem for this card, but I don't know what other problems would arise to other cards by doing so.

Any better ideas?

Did you try add blacklist.conf to /etc/modprobe.d  ?
or
Every module you may need load on boot:
/etc/rcS.conf LOAD_MODULES=

Hmm...

I just had a look at /etc/modprobe.conf, /etc/modprobe.d/ and blacklist.conf

It seems that the 'blacklist' only is used by udev. Modprobe seems always to load the modul, regardless of any entry in any blacklist.

I also had a quick peek into the Debian manpages:

-b --use-blacklist
                 This option causes modprobe to apply the blacklist commands in the configuration files (if any) to module names as well.  It is usually used by udev(7).

Unfortunally, modprobe in the busybox doesn't have this switch. :-(

If I preload the modul with "LOAD_MODULES=" then it will be loaded, regardless if the card is inserted or not, and it doesn't prevent prism54 to beloaded by a later 'tazhw detect-pci'.

It seems there are two different cards with different chipssets out there which report the same vendor and product id.
Therefore I guess there is no 'clean' solution to this problem but to find out which one you have and to remove the other module. :-(