I got most of this from literature available on the Internet, so take it with a grain of salt. I have no experience measuring noise but I played a noise engineer on TV.
(If you're not old like me, that joke will go over your head.) I believe that the best op amps are just below 1nv/vHz, but you have to also check the specs for 1/f noise (some are better than others), and their current noise limits them to low source resistances. Chopper-stabilized amps have no 1/f noise, but their wideband noise performance is relatively poor.
Even at 1nv/vHz, a 1MHz bandwidth will give you 1uv of wideband input-referenced noise, so it looks to me like an amplifier is going to be a challenge. As you suggested, you might be able to RSS the amplifier noise out with a caibration scheme, but that means your amplifier noise really needs to be "white" (I think).
However, you gotta ask yourself, "How do they measure 1nv/vHz?