Failure to probe atapi CDROM unless booting from it

[John]

When it is identified, the device is identified as follows:
Jan 11 17:56:56 pearl /kernel: acd0: CDROM <Compaq CRN-8241B> at ata0-slave PIO4

That only happens, however, when I boot from said device (this is
on a Compaq Armada M700, P650, 320Mb RAM, BIOS 686H 3/15/2002(latest)).
Otherwise, I get
Jan 11 18:09:55 pearl /kernel: ata0-slave: ATA identify retries exceeded

I am running 4.8R, but I've looked at ata-all.c from 4S, and it doesn't
look significantly different with respect to this section of code.
Well, it doesn't look at all different with respect to the section of
code that produces the error messages above.

So - is the BIOS doing something to kick-start that drive that should
be added to our code?  Anyone else got a similar drive which is
working correctly?  I've heard from other M700 owners, but they
haven't identified their drives.  If other drives are working, and
no-one else has one of these, maybe I can arrange to trade with
the company I bought this from for a different drive, or just buy
a DVD-capable drive from them and be done with it.

Based on this experience, however, I'd like to be able to specify
exactly which drive they should send me, since, apparently, not all
drives are quite interchangeable.

Other suggestions?
-- 

John Lind
john at starfire.MN.ORG