Spindle speed does jeopardize performance and reliability due to excessive vibration - unless you counteract it appropriately. That's why RAID edition, NAS and enterprise drives have special vibration compensation built in. If you don't need to do this due to lower spindle speeds you can offer cheaper drives.
And we seem to have a different opinion about how cold cold data actually is. For me this is infrequently accessed data - otherwise it wouldn't be cold. It's the last randomly accessible storage tier, right before flushing to tape (if available). This could be archives, backups, unimportant media files or whatever. So why care about the performance of things you almost never need?
A quick google pretty much reinforces my point. They're even saying shingled magnetic recording disks are ideal for this task!
Of course in pure 512/4k random access no HDD could saturate USB 2. But why would you access your cold data like this? If this ever happens you're extremely unlucky or you've put your data into the wrong storage tier. It should at least have been on SSDs, as even 15k HDDs are far too slow for such cases.
What does "mutually exclusive" mean if not "if I want to increase one of them, the other one has to go down"? Areal density is the same as density per platter (what I refer to as platter density, not the number of platters) as long as we're talking about the same physical HDD size (3.5"). And it obviously depends on recording technology - for any spindle speed.
Honestly I don't know why you're bringing up the Raptor here. It just reinforces my point: despite running at 10k rpm it "only" achieves as much sequential throughput as a current 7.2k rpm Seagate. That means its linear data density must be ~30% lower. And they're not doing this just for fun, it's simply because at higher rpms you can not reach the same density (for a given recording / head technology).
Actually, there's also the factor of disk diameter, which gives the 3.5" drive an advantage of approximately a factor of 1.4 over the 2.5" Raptor. Factoring this in both drives would be about tied for areal density. But note that with lower diameter platters vibration (again!) is reduced, which makes it easier to achieve higher platter densities. Which is why the Raptor transitioned to 2.5" in the first place.
After my previous comments it's probably no surprise that my simple answer would be: because different markets are better suited for different spindle speeds, and sometimes other features (like that vibration compensation). However, there's something I really don't like: separate product lines for drives which only differ in firmware. Not sure how common this really is, as they could be subtile hardware differences.
But there are 5.4/5.9k rpm HDDs being sold specifically for digital recording. I can not imagine any hardware modification neccessary for this, hence they should just make the firmware tuning for these apps user selectable and kill off that product line completely.