512 bit wouldn't help against quantum; the problem is Shor's algo is not exponential in that parameter, it's polynomial so it basically doesn't help.

Meanwhile we're at about 128 bit security (think birthday) w/ 256 bit curves, at least as far as anyone knows, shifting to 512->256 bit security is obviously "better" but otoh nobody seriously considers anything above 90 bits is attackable (and remember that scale is exponential!)