A lot of the "format" for many storage systems is effectively encoded in the hardware (e.g. the speed of a stepper motor, or some thresholding in a signal filter), so short of using, say, an electron microscope there's no way to even read the data in some raw format without the right hardware. At that point it may very well be economically infeasible to recover the data, and the physical storage might degrade before the feasibility reaches a tipping point.

Storing digital copies indefinitely is somewhat of a more modern invention. The digital copy still has to be stored somewhere, and that somewhere used to age out pretty quickly; CDs were seen as archival, but that's becoming less practical. Nowadays you can keep copies in cloud services, and as long as you keep copying them around then it's probably easier to have long-term retention, but you need to keep doing that indefinitely.

That doesn't help when the bits have literally fallen off the tapes.

That is an argument for better digital storage. Not cumbersome and expensive dead tree storage.

I don't think it has to be one or the other. Why not both?