1) Remove any limiting or maximizing from your master fader. Mild or subtle compression is OK but avoid any makeup gain that is smashing into the digital ceiling as that cannot be undone. Turning the mix down is easy, undoing the compression/limiting is basically impossible, the damage is done. For example, you can crank the output gain of Waves Renaissance Compressor
and it won't clip, but it acts like a limiter when you hit digital zero which again, can't be undone. If you must use compression on your mix, avoid extreme makeup gain like I just described. If this is part of your sound and you don't want to remove it, your mastering will surely be done "in the box" and probably not come back to you much louder, or at all.
It's understandable that you might have to do some faux mastering to make the client happy during mixing, or to get an idea of what will happen, but I always suggest removing that stuff for the final bounces for mastering.
When I get mixes that are around -15 RMS or hotter, it makes it pretty much impossible to run through the analog chain without sounding like shit. Submitting loud mixes for mastering usually means your mastering engineer can do less. If you're doing a vinyl release, it's especially important to leave plenty of headroom. There's no magic number for headroom, just don't limit or over-compress. You could brickwall a mix, and then turn it down -6dB and claim you have 6dB headroom, but it's useless if it's smashed and the dynamics are gone.
2) Noise. If there is buzz, hum or hiss from something in or on the mix and you'd like it to be removed or lessened, be sure to leave a nice sample of JUST the noise before or after the song so it can be used to remove the noise with RX3 or something similar. I seem to get a lot mixes where the engineer will be so worried about the noise that they crop it out very tightly, which leaves no clean noise sample to work with. The mastering engineer can quite easily trim heads and tails when needed, so it's best to leave those noise profiles before and/or after songs if you'd like them reduced.
For example: Your song has a guitar intro and the amp is very hissy, like a classic Fender tube amp. You trim the guitar tracks in your mix session so there is no hiss before the first note. Once the song is mastered and much louder, the hiss that is underneath the guitar intro is more noticeable and bothers you, but since there was no sample of just the hiss with no guitar playing, it makes it nearly impossible to remove hiss properly. All it would take is a 2 second sample of only the noise which is usually found before or after the take, or maybe in the middle if the player stops playing for a bit.
I always do tight crops after I capture back through the analog chain, and after I do any noise reduction. I know it might be your instinct to try and crop all noises out ahead of time, but it doesn't help if noise reduction is to be used. If you need to keep it cropped tight, just paste a bit of the noise from the noisy tracks after the song is over.
Sometimes the noise might not even be noticeable, or an issue until after the song is mastered up loud, so listen close for noise that could be problematic after mastering.