Your experience in real world listening seems to run parallel to mine and also to what I have been seeing with the Clover QC unit. Yep, diff. media and diff speeds, burners and so on do sound different to me too.
But the gold seems to be about the "roundest wheel". The MAM-A silver, well the quality (listening and Clover testing) has been all over the map from lot to lot. Sometimes very good, but sometimes, really bad. Many of the same problems you mentioned, partic. collapse of the stereo image. The Taiyo Yuden media (80s only) has been more stable lately (testing both very subjectively and with the Clover too) so have been using that for both CDROM work and often for client refs too. So far, no one has kicked back a ref as not playable though it's an 80 min blank, so I guess most clients are using newer players. My car system (late 90s Jeep) will eject any 80 minute CD put into it, though it plays 74's fine.
I've had REALLY bad luck with the MAM 80 blanks though -- both silver and very surprisingly, the gold ones too. Not only do they subjectively not sound as good (same program, same burner) as the 74's but the Clover thing generally shows them to have many more errors. In fairness I think all I ever bought of the 80s was from one batch though. I usually buy from Media Supply in PA, a big MAM-A distributor.
The Clover unit is fairly cool for what it is - consists of a graded and modified Plextor drive (SCSI) which allows it to measure a big variety of error rates, and will even let you look at the pit geometry if you hook up a scope to it.
It requires a COMM port to run besides the SCSI card (they include an Adaptec controller), but will run on just about anything. I have it set up on an old DEC Pentium 200, one of a very few packaged computers I've ever owned bought as a closeout when Compaq bought DEC.
Clover's info claims that some of the errors it is measuring are likely causes of some of the same things you and I are hearing/seeing-- it measures much more in terms of errors than just BLER.
But like anything else, it is not totally empirical either-- since there are the variables of the computer host, also many others. You check to see that the thing is still calibrated fine and that all is well using 2 test discs they supply with it -- 1 known good (they give you the analysis of the disc when the unit was shipped) and one with def. known errors (they give you that analysis too). So far, our system seems very repeatable to the standards from when it was shipped (about 1.5 years ago). But seems the more bare bones the host machine, the better. I have it on a very simple Win98 system that is about as vanilla as you could have, no Internet etc either. I just use that computer for QC'ing, really nothing else except printing out the documentation.
That's also why I check masters twice with it-- the results always will differ, but only very slightly with a good master, and the results will be fairly close to the "standard" cal. disc when all is well. I print the results and ship with the masters too.
www.cloversystems.com if you are interested at all, but it isn't cheap, though about the cheapest thing beyond toy class that I could find when I was looking around. The owner apparently did some work way back for Dave and MTU in its very early days, though last time I spoke with Dave he didn't remember the guy -- that was some time ago however. They make a few different types of units, the one we have is called CDX.
Seen the same thing with DVDs used for data backups. No surprise that cheapo media is not reliable -- but the one brand that seems to work well time and again for those has been Verbatim. We have many DVD drives, from 4x to 16x, mostly use them for data backup which is always fully verified after writing (as a separate verify pass, using Backup Exec). The Verbatims seem to be very reliable, most others, not. Will have to see how they hold up over time as storage media-- but I did restore a project done a few years ago recently that was backed up to DVD and it worked out fine. With really critical stuff that is very likely to be reworked years later, I back up to both DVDs and DDS data Dats which is what I used before the DVDs, and usually run 2 sets of DVDs too just in case. I get paranoid that way!