He didn't claim it is broken. Only that it can be broken 2x faster than before. RSA 4096 as recommended by the FSF is still secure. RSA 2048 might be breakable by the NSA. But so far we are at 800-1000 at risk.
Yes, because RSA 2048 is broken, not RSA 4096. Almost everything uses 2k, nobody uses 4k. The cryptosystem infrastructure would be broken. Not talking about the possible backdoors in the ECC NIST curves. Thanksfully we have the djb curves there.
But then there's that line: "This destroys the RSA cryptosystem" in the abstract of the paper.