I heard I also need to keep the angular alignment between platters pretty much perfect, at least on drives with servo surface. Is that really the case?

Yes, generally the low level formatting* is only on one head on modern disks, and since at least parts of the firmware (more now than ever) are also on the disk pack, and rom space is still more expensive, they aren't designed to work in such conditions

* modern disks, as in newer than the literal definition of "MFM/RLL with integrated controller", "the ones you can't LLF at home", can in fact (depending on the brand and model series) be... hmm, I'll call it medium level formatted (can't recreate the factory alignment, but can do a destructive self test and properly exclude bad heads/sectors without consuming the spare area for remapping of future bad sectors), the generic name for this should be "self scan"

Electronics swap - most modern drives (different definition than above, but still give or take "mid 90s") have "adaptables" in flash memory which means most of the time you can't do a proper job, or even get away with it, unless you also transfer the appropriate parts of the eeprom (it's not just about firmware but rather configuration, as in different models of head preamp and per-unit fine calibration), which implies that:
1- if those are corrupted you're usually screwed (I assume you can't afford hard professional gear like the famous PC-3000 [starts at 6 thousands] but even that doesn't do miracles, effort is required in these cases)
2- most problems can't be fixed in this way (if your head/motor controller is dead chances are it's CAUSED by a totaled head preamp, etc)

If you want to learn more, hddoracle is the go-to site, two days worth of reading and that's IMO the most educative and relatively unbiased one 😀

---

High level surface scans - generally reads of bad sectors do not cause remaps (they cause "pending sectors" in SMART), only writes will (usually only if the write actually fails - it is possible on many drives to create deliberately bad sectors with "long write*" commands, since the platter is still fine it will fix itself properly on the next normal write), that's why MHDD and Victoria have an "erase delays" feature that really means "write to slow-to-read/unreadable sectors to recommend the disk's firmware to remap them"

* basically the "raw DAO 94" of magnetic disks, you can freely write even the ECC with that

And of course there are different definitions of bad blocks at play there - the bad cluster markers in FAT, NTFS, EXT?, etc come from the days of non-remapping disks* (or modern ones once all the spare sectors are in use) and you won't see Scandisk, badblocks, or the like detect any until that point; remapping is internal to the drive* and transparent to the computer apart from possibly in SMART values...

* Controllers for "the disks you can LLF at home", instead, usually ask you to input the list of cylinders to skip when you do that, but the result is just a quicker fail AFAIK (a BBK error that will be picked up by scandisk's surface test and turned into a FAT bad cluster), not that track being skipped internally

Thanks, I'll visit hddoracle when I have some proper free time for reading. Wren III Design 1985 document is full of interesting details as well. I'm aware it's not possible to format the servo surface on HDDs, that requires equipment that the factory had (probably also a drifferent head, the one in HDD is usually read-only). But I've seen some old IDE (and not only just the "RLL with internal controller") actually do something when low-level format track command is issued. Data on HDD was lost and at least some bad sectors went away. I wonder how the later HDDs with self-servo tracks were made - was the head assembly inside the HDD used to write the original positioning data and sector headers? If so some disks might have hidden FW command to redo it (or refresh it), assuming this wasn't done with external electronics that had better voicecoil control.

Regarding my own question about sticky HDDs, sadly the subject of my experiment died before I could fully test a PTFE based lubricant. One of the heads broke off during my attempt to spin the platters by hand (to apply the lubricant). I've been able to start the disk that way a dozen times but somehow this attempt failed. I should've probably given the HDD a few taps on the side first, I'll be doing that from now on - if I get another terminal stage patient. All I've discovered is the PTFE lubricant does improve the "spinability" of the platters a lot, but I do not know for how long, or if and how badly it'll affect R/W operations.

Seagate ST10 series (like the U 5 I have) seems designed to get the "true LLF" done after most of the product is assembled, there are service commands explicitly designed to interface with the servowriter, though I can't imagine how would it work (no cover and a second set of heads sneaking in from... somewhere? or does the drive write the sector headers and this "servowriter" just spins the disk at a calibrated speed?)

As for what torture can a disk get away with... I was burned by a seagate rosewood (spins but no SATA communication, don't have 2mm dupont cable for the serial port, you can polish a turd but it remains one, etc) so I removed the cover and put various sprays/liquids/screws on it, none did any visible macroscopic damage, had to put a sticker on the platter to get a head to fly off 😁