Many home computers in the 1970s and 1980s had two floppy disk drives
Home computers had any number of drives from zero to 6 or 8, all depending on budget. Even rather restricted systems like a C64 had by default addresses reserved for 4 drives.
However, disk drives were not cheap.
And price is why there were single drives. Home computers are all about getting a low price point
Given that it was rare that a system or program or game would actually need to read from both disk drives at the same time
It was extremely common to have programs using two or more drives. One holding the program disk, so overlays could be loaded according what was to be done (remember, there were no multi megabyte memories holding everything), plus a data disk holding your document(s). Of course there's always the chance to swap disks whenever access to either was needed.
In fact, it's the very reason why the DOS on the IBM-PC emulates a drive B if there is only a single drive available, by managing them as logical drives and checking disk IDs with every access. Doing so allows to handle two different disks without interaction from the program, while avoiding overwriting the wrong one.
except for disk copy tools, and even those can easily use the RAM as a buffer)
I have been wondering lately if a "jukebox" approach would have been feasible.
Sure, and they have been built and offered.
I assume the read-head was the most sophisticated and expensive component of disk drives.
Not really, especially early on the mechanicals were the most expensive part. This is also the reason why Processor Technology designed their Helios II drive around he Persci 270 drive - a drive having two disk slots, each with two heads, but only one frame, one drive motor and one stepper motor. The idea was to offer the ability of two drives but cutting cost. Would have worked fine, except the floppy became such a success that, by sheer scale of volume, the price for two single drives fell below what the otherwise simpler mechanicals for the Persci 270 would cost.
Another example was given by Apple. They only bought the bare mechanism and head mount, but designed everything else at a lower price themself.
Sure it would require some sort of quite sophisticated mechanics and controller logic,
Controller wouldn't be a big thing, it's the mechanics. it would need at least another motor, able to stop exactly at multiple positions and rather fine mechanisms to move the disk pack as well as the disks in and out. Such a setup would be way more expensive than a second drive, while at the same time not being able to access two disks concurrently - at least not without several seconds between each change.
So did such devices actually exist, would they have been feasible, and should I get a time machine and invent those? :)
No need to warm the Flux Capacitor, such drives have been made, but not for home computers, but professional systems. In fact, they have been around longer than home computers. IBM offered in 1973 the IBM 3540 Diskette Input/Output Unit for their mainframes which could operate up to 20 8" diskettes at once per drive and two drives in maximum configuration (B2). Then again, floppies were already back then way too slow, so IBM had as well IBM 3747 Data Converter, which could autofeed floppies to be copied to tape, which then was transported to the mainframe - they were part of the 3740 Data Entry System.
Similar attempts have been made in the 1980s for 5.25 drives, but were never really successful. It was simply less expensive to go ahead and buy an HDD instead - or, for most home-computer users to go ahead and buy a drive with higher capacity.
The only niche market for auto loader systems were disk copy stations used to manufacture software. But their purpose was to feed a stack of disks one by one, not to select between different disks of that stack. Equally important, they were way outside the price range of home computer users (several thousand USD per station), often self contained and usually HDD based.
An early example for such a mechanic is documented in this patent application of 1982. It feeds a stack of disks from the top and and selects the output into two bins below. Most likely to separate successful writes from failures detected in verify. Some other similar patents from the 1980s
An exception to these production orientated devices might have been some auto loaders for the Mac. The most well known (in the US) maybe the Jukebox Five Automated Disk Changer (video). It allowed to sequence thru a stack of disks, accessing them in sequence, but not random. Useful for installing software, or small scale production. It wasn't software controlled, but relied on the auto eject function of the Mac. whenever a disk was spilled, it was dropped below and the next in line was inserted. The handler was introduced at the 1991 MacWorld Expo by Fifth Generation Systems, a company focusing on backup solution - which benefited a lot from such a device. There were other similar systems as well
The only disk technology where changer systems had a tiny niche with end users were CD changers during the mid 1990s. Like the well renowned Pioneer 602X/604X/624X series, or the later priced Nakamichi MJ-4.8s (4 disks) and MCD-1020 (7 disks). Still, they cost >4 times the price of a single drive professional (read, already expensive) drive. Here again it was a matter of scale - in this case the fact that these drives were originally designed to be used in high end car audio systems, which produced some volume.
Of course stuff like that gets reinvented over and over again, so here a video of a nice arduino based hack.