Does seem odd, considering "bits is bits" as it were.

However, an optical medium provides no electrical connection between
the devices and hence isolates them electrically.

Coax does not provide electrical isolation.

This difference makes me wonder whether a ground loop might be
interfering with reliable digital communication over a co-ax link. Of
course a piece of on-board hardware that drives the co-ax but isn't
involved in driving the fiber might also be to blame.

Are both ends of this chain plugged into the same properly grounded
electrical circuit? Are other devices connected to other circuits
electrically connected to these systems in any way?

I considered that as well. And to answer your question, the equipment is
installed in a car, where ground loops tend to be prevalent. The computer,
DAC, and amplifier grounds are all in the same location; there's no
"alternator whine", which is often indicative of ground loop issues (or poor
grounding in general); the computer's chassis is grounded to this point and
so are the rack rails. But I would think that if it was a grounding issue,
and there was some sort of interference introduced into the signal, then I'd
be having trouble with the digital transmission and it would "cut out" - all
or none, I guess. I would think that the problem must then lie either with
the digital to analog conversion or "after", or "before" the ADC? But is
ADC even employed in the computer? Sure, there's an analog sound output
onboard, but why would it have to go ADC-DAC to give me a spdif output?