[2] is an awesome thought — that a nearly perfect sphere might be required for some calculation; it seems reasonable as a requirement — but isn’t, actually. And that Shor’s algorithm might have a similar impracticality we don’t presently appreciate.

Reminds me of this way in which laser interference patterns can be used to find primes. But, like many analog computing situations, it requires a degree of sensitivity to signal vs noise that is essentially impossible.