Laser-guided SuperDisks got up to 240MB, and Zip drives with floating heads got up to 750MB. At that point you're basically operating a hard drive without a dust seal, so it would be pretty hard to go much further even with modern tech.
But if you did figure it out, let's use the data density of LTO. If we estimate a 3.5" floppy as 50 square centimeters, then that's the same area as 40 centimeters of tape, which fits about 7GB on LTO-9. A bit more if you can get double-sided writes to happen and not interfere with each other. Less if you have the dead zones of an actual floppy.
A ~5GB floppy disk actually sounds like a pretty useful thing at first, until you realize that you can get USB thumb drives that size almost for free these days. At least the floppy disk would probably be a reasonably secure way to store data for decades, unlike the flash drive, as long as the disk (and the drive) were made with 1980s/early-90s quality and not late-1990s/2000s quality (floppy disks in the 80s were supremely reliable, whereas ones in the late 90s were hilariously unreliable).
If you took some modern flash and ran it at half capacity by using 2 bits per cell, gratuitous error correction, and lots of spare blocks, I would expect it to store data for decades.
But if we're positing reasonably-priced drives with that kind of tech, I would rather it actually be a tape drive. Let me get a 9TB tape for $40 or less, even if it's only useful for backups and video files.
Beyond about ~100 MB historically speaking the industry shifted towards magneto-optical systems. They do store data magnetically - but they're read with lasers, since tracking and physical wear and tear from a head in contact with the medium, were the major limitations when it came to increasing floppy disk density. There's a neat effect where certain kinds of crystals vary their reflectivity depending on magnetization, which is how they work.
Sony's MiniDisc was an example but in the '80s and '90s there were many competing variations.
I looked it up later yesterday, and it looks like after the Jaz drive we tried out a design where the read heads were built into the cartridge. Presumably to avoid the problem with the relative size of dust versus the flying height and of a byte on disk.
But we also started seeing triple platter or sometimes more in hard drives, making the gap even bigger. 1.8” hard drives in a hot swap slot would work pretty well for something like the device in this article.
I would take a future bet that the moment we have something that can replace flash memory we will drop it like a hot rock and then tease each other about how dumb we were to entrust real data to being stored on it.
It's not like we have a choice. We don't have any robust storage technologies for normal use, so it's flash and hard drives being unreliable versus not storing data at all. Tape is the best shot for large medium-term archives, but I bet if you baby flash with a small number of writes and cool storage temperatures it'll also last 20-30 years.
