Agreed -- the price contrast between the unit itself and the component replacements and upgrades is a turn-off, however going into a ~30 year vintage pro audio sampler thinking it will have a low cost of ownership or repair feels unfair.
Yes, the base unit themselves might be cheap. But once you're looking at enthusiast-produced LCD replacements or optionally-purchased-in-the-90s expansion cards, having sticker shock is expected.
LCD parts are cheap if you want to put the domain engineering work and solve engineering challenges yourself.
They are expensive if you're one-stop buying a targeted kit and just trying to max out your upgrades.
An HN poster might be willing to put in significantly more time head-desking and soldering to save $200. A producer buying a vintage Akai is not necessarily looking for an EE project.
To that producer, the functionally upgraded unit is "worth" a lot more than $300 since they're selling the products of it.
They are cheap to find -- on Yahoo Auctions Japan I got an s3000xl for ~$80 since the 8-out works in the MPC2k/xl and go for as much as ~$400 alone.
But that doesn't mean anything about the cost to repair, upgrade, or in the case of LCD replacement, modernize. The memory, flash, effects, and expansion cards are highly sought after, and you're competing with bigger budget established producers.
As noted, the expensive MPC60/3k replacements are just generic T6963C / RA6963 displays with the circuit bypassed (since sampler has the LC7981) and the correct pitch ribbon cable soldered.
I did a backlight replacement on a S3000XL, a LCD replacement on S900, and found a cheap ($25?) LCD replacement for ASQ10 (same as MPC60/MPC3000).
Jazzcat + Ebay sell expensive replacements, but they are actually just cheap LCD displays with the driver circuit bypassed (since the MPC/sampler has a discreet IC) and an epoxy blob to prevent copycat work.
I need to dig up the wiring diagram, but the gist was from an EEVBlog forum post and only took an evening to reverse engineer.
NextBSD ("FreeBSD X") operates in a similar space and was also looking promising, but it looks like development stopped.
In that approach, the BSD kernel was extended to support some Mach syscalls and libdispatch, launchd, and a few other userspace macOS-isms stuck around too.
Honestly, I'd probably just pay the premium and buy a kit if I did it today.
I did this during COVID lockdown when my hobby/project time was valued differently.