This is for both actual examples (e.g. Conner's "Chinook", which had two complete head assemblies), and hypotheses (if you think of something that could work, but don't know if it actually will).
Maybe there is a way to use fluid bearings to support the head assembly??? I know fluid-dynamic bearings can't directly replace ball bearings in the head assembly, as FDBs rely on rotation to circulate the fluid and backwards rotation can actually expel the fluid. (The fluid is magnetically retained with the spindle stopped, so it doesn't leak even though the bearing shaft is not sealed.) This is why the head assembly is supported by BBs even today. But my idea is, if you put a small FDB-equipped spindle below the head assembly and used a second FDB between that spindle and the headstack, would that work correctly and allow drives to sustain heavy seeking for a longer period of time???
In essence: First FDB's casing fitted to the baseplate - shaft connected to the additional spinning motor and the second FDB - second FDB's casing fitted to the headstack.
Obviously this would require an extra group of 3 or 4 terminals to the PCB (depending on whether the motor windings are delta or wye connected, assuming the motor is 3-phase like practically all HDD spindles). This system would also mean there would be no easy way to secure the top cover of the drive to the bearing assembly. (Then again, most FDB spindle motors are not secured to the top cover (there are exceptions, but usually only in high-end drives).)
Another thing I thought about was somewhat along the lines of the ancient SCSI Barracuda 2HP, which was actually capable of using two R/W heads at a time, across both sides of a platter. (Other drives are one-at-a-time.) Modern areal densities would make it extremely difficult to line the heads up with each other (drives actually have to seek when switching heads, as they can be multiple tracks from aligned), though, so my idea instead has two heads (actually four in total - 2 read and 2 write, as magnetoresistive technology requires separate read and write heads), approximately a track width apart, accessing adjacent tracks, on each slider. Therefore, the drive would seek by 2 tracks at a time when sequentially transferring data.
There are actually HDDs out there (some WDs and the infamous Maxtor DiamondMax Plus 8) that use(d) a spiral layout to (nearly) eliminate track-to-track seeking, theoretically increasing the STR, but would only work well if you only switched heads between zones (not applicable of course for DM+8 which was single-head only).
