Sometimes nothing, but usually an SSL bus comp. If it serves the track and feels right, why not keep it there.

I'll always suggest to leave the mastering when it's time to master.

It might be a good exercise to help you realize what you're doing or not doing to get the desired results.
In which case, you might have to look outside of the daw and into the room and equipment. Other wise, consider specific techniques on mixing a certain component or string of components that effect the element in question,that best gets the results you want.

Often times, You can do things individually that you've done on a master that might work for that singular element. This logic stems from the focus and pointed question you have "oh because i want this to happen to X and i did on my master, it must be ok". But the focus is on X, so shouldn't you just do that for X instead of entirely having it for the master?

Plus if you treat individual elements, it gives the master track more life in it's entirety.

This is just going from Macro perspective to Micro perspective. You need both, but understanding which perspective to use where is the key part.

I think it's best to start out with a clean master channel - though I advocate system-wide analysis plugins just for ease of use. I have a spectral analyzer on my master chain, and I'm about to add something I got from the EDMProd.com Start to Finish videos on Progressive House - A high-cut filter set at somewhere around 70-100 Hz that I can turn on and off while I'm mixing to help check the low end.

I'm definitely for a clean master as well.
The worst it can get is when you try to fix a specific problem from two sides, where one side is the specific channel and the other one is the master...Just that with the master you're affecting all the other elements in the mix too. This is my opinion though, after I've run into making this "mistake" myself.

I master in a completely different project. Simply because all my mastering plugins chew up my CPU. So I will finish my mixdown, Export to WAV and insert into a new project.
As far as you said with the TP basslane, I prefer ozones imager.
My chain is normally:
Compressor - usually the glue or fab filter pro -C
Ozone: usually using EQ (mid / side), Exciter, Sometimes the multiband dynamics if needed, the stereo imager (generally making lows more narrow and spreading the tops).
EQ - usually fab filter 2
And limiter - Usually Oxford - I just really like the Perceived loudness you can achieve there. The limiter seems to be really Transparent and can be pushed pretty hard.

Sometimes I use the PSP vintage warmer as my compressor depending on the track but really sparingly.

Although I do not master as I go,  I do have an idea of how I am going to approach mastering whilst creating the track. I kinda start off with the same mastering template during all of my mastering.

I'll throw on a Freq Analyzer to see what's happening level wise with the mix. Then I'll do a "rough" master with a multiband compressor and maximizer to see what the mix sounds like after being compressed and limited, but I turn those off when I'm actually mixing.

Yep! Frequency Analyzer and Stereo Imager

I think it's a bad idea to have anything even slightly processing the master while working on the track. Here's a reason case by case:

- EQ: Changes phase response for the whole song. More steep filter you use, the worse. Not so much of an issue, but in extreme cases it can add a wierd effect you get used to, but listener won't.
- Linear phase EQ: Adds delay, resulting in input lag. As the processed channel is master, no latency compensation will help you.
- Clean compressor, heavy settings: Screw's your idea about transients that actually happen in the track.
- Clean compressor, light settings: Not much effect as it only cuts highest transients.
- "Character" compressor: Adds saturation that potentionally might fool you into thinking you have a mixing issue somewhere.
- Harmonic exciter, tape sim: ...same reason.
- Limiter, heavy setting: Same problem as with heavy compression. Your idea of transients is screwed. You'll mix the track to liking, mastering engineer WILL ask you to take the limiter off and then you'll find out it doesn't sound good anymore.
- Limiter, light setting: You might miss some really high peaks. Taking it off and mixing quiter actually saves you some hustle in that case. Master limiter will bring it up anyway.
- Stereo widener: You get used to unusually WIIIIDEEEE sound. Your stereo image will fall apart when for example mastering engineer dials it back a bit. (Or when it gets played in mono. Cheaper cell phone speakers, some clubs...)
- Bass narrowing plugin (TP Basslane): It cuts low frequencies out of the side signal. If it makes a big change, it suggests there's a mixing problem somewhere else (not enough low-cutting, too wide bass) you're not hearing, thus not solving.
- Reverb - Some people like to "test" their track by putting it in a "virtual" room. It can help as a check for low frequency room behaviour, but shouldn't be on the whole time. It screw's your idea about how your ambience is behaving.
- Room simulator - "BUY THIS AND YOU'LL HAVE MONITOR SOUND ON YOUR HEADPHONES!" ... yeah right. Dunno. Tried, wasn't convinced. ...but whatever it does, I'd be really carefull to make mixing descisions upon it.

Analysis stuff like frequency analyser or steroscope is ok. It doesn't process the signal.

...also it's just an opinion. If you put stuff on a master and it works for you, ignore this.