[geofft]

The article claims that a well-known person in Linux kernel etc. communities has a 4096-bit RSA private key with two "factors", one of which is 231 (which is itself not prime).

Standard RSA involves a private key which is the product of two prime numbers, and when we talk about "4096-bit RSA", we generally mean that the private key is the product of two 2048-bit numbers. Due to randomness, they might actually be 2047- or 2049-bit or something, but having one 8-bit prime and one 4088-bit prime -- let alone one 8-bit composite number -- is generally not what we mean by 4096-bit RSA, nor what software is intended to generate. It's well-known that the strength of an RSA key is restricted to the size of its smallest prime factor. (When I TA'd a crypto class a few years ago, I gave a homework assignment to break a "512-bit" RSA key that was intentionally generated as a 50-bit prime and a 462-bit prime: the entire class was able to factor it without even thinking about hardware beyond their personal laptops.)

So the mystery is not so much how it was factored, but how the key was generated that way in the first place, and whether this was even an intentional part of this person's PGP key. Other comments here imply that this particular subkey isn't even usable, so the answer to the second part seems like "not really".