Where is that measurement been taken from?
Your scope pic looks to me to be more AC line ripple than noise. If it is ripple, changing bjts is not going to fix it.
Noise measurements are better taken with a audio analyzer, so that you can apply filters to differentiate noise vs ripple.
I would suggest to increase the resistance of R41,42 to say 100 ohms, as another option, if you want to reduce the opamp supplies a bit, it would also help in reducing ripple, so would beefing up some of those ecaps, like C36,37
You have to watch when you replace those slow opamps with higher GBP opamps as they can oscillate on you. Applying local decoupling, using 100n is always a good idea and recommended. If you read the new TI NE5532 datasheet, it shows the recommended way to decouple the devices.
I guess 200mV offset is starting to get higher than one would like, but it you do not get any speaker popping I guess that you are okay.
C19-22 could be increased a fair bit more as well. See what fits, could go up to 1000uF, but starts to get expensive