Rust is still a problem with those little ferrite rings, though.

Depending on the environment you maintain for this device, it can actually be pretty delicate. I have some old (~25 year-old) electric motors, where the magnetic properties of the bigger ferrite rings have been degraded by rust. I guess it probably depends of the quality of the piece.

This means there could be one tiny rusty ring in one of the center planes, and that would effectively result in a bad block that exhibits inconsistent behavior for that single bit. A person would have to test for that. It could be programmed around, and you could still have a fully functional object, minus one byte or word (or bit, depending on how you approach it).

You'd also want to keep it free of dust, since dust CAN be metallic, magnetic and conductive. The item for sale looks like it's been taken care of nicely!

From wiki: "A typical magnetic region on a hard-disk platter (as of 2006) is about 200–250 nanometers wide (in the radial direction of the platter) and extends about 25–30 nanometers in the down-track direction". It's not trivial to say "Let's get X and miniaturize it to this scale".

Flash memory uses standard semiconductor techniques, and HDDs simply employ platters with tightly controlled properties.

Core memory is non-volatile. If you hooked it up to a computer, you could read out what was written into it. Theoretically.

NASA recovered the Challenger's core memory out of the ocean and was able to read it out months after the disaster.

"On March 19, 1986, NASA announced that four of five Challenger General Purpose Computers (GPC) had been recovered from the Atlantic and moved to the IBM Federal Systems Division facility in Owego, NY. The GPCs were cleaned under controlled conditions and submerged in deionized water at Kennedy Space Center prior to air shipment March 16, 1986, to Owego. The GPC ferrite core memories were examined for any possible residual data -- a process that at the time was expected to take several months. This information was in the form of data--not onboard voice--and this path was pursued to add any possible additional information to the accident investigation. Many weeks later, it was found that the additional data frames did not measurably add to the information already gathered during the investigation."

Ref: http://www.aerospaceweb.org/question/conspiracy/q0258a.shtml

Iron core memory, so it would probably still read/write just fine, if not, fixing it shouldn't be too hard if you've got the hands for it.

Reminds me of this core memory shield that was developed for the Arduino: http://www.corememoryshield.com/