I am not familiar with Wavelab so don't know how different v5 and v6 are so the following may be spot on or completely different. The ts reference section has a tutorial for matrixing two sources in Wavelab5.
That tutorial does not discuss time stretch but rather syncing, which does not solve the question of timing issues. I've found that if there is drift you may need to cut the files up into smaller segments and may even need to remove very small segments of the file if the problem occurs during a song.
I've never really figured out how to use the time stretch, which would be really useful for analog sources. This is something I need to learn. anyone have a reference or answer?
You're not alone. I've found my success rate in using the time-strech function to be < 10% (and that's being generous) and the issue is that each time-stretch operation can require 20-40 minutes or longer. I don't believe that WL6 has much in the way of support for newer hardware (i.e. quad-core processors), so it's possible this operation runs much quicker on a new version, but it's slow as heck in WL6.
Anyways... what usually occurs is that it takes me 4-5 tries, at least, to get the right amount of stretch. I probably just suck at math (I do) because I am seemingly off every time and it turns into a routine of trial-and-error.
And then there is the changes in drift over time as you mention. I've seen plenty of recordings lined up perfectly at the beginning and end, but seemingly very far apart in the middle. The rate of a word clock is not an absolute constant and it's definitely possible to see large variations in drift over seemingly short time-segments. When this happens, the only option is to start chopping up one of the files and sync in small portions. bleck.
Quite frankly, I don't think any FAQ or Guide is going to help anyone attempting this process and fair warning... if you are going to attempt this be prepared to spend a lot of hours trying to get it right with the possibility that the only solution is the 'chop' method.