Gain matching makes more sense to me than level matching although at first I didn't really get the advantages. After I let it synch in a bit it just makes a ton of sense over level matching IMO. Another good thread on it here http://www.avsforum.com/t/1282064/two-subs-gain-matching-vs-level-matching
Post #5 is the process but lots of good discussion there.
On the phase controls I think it is good to let Audyssey come up with the distance for the further sub which may actually be different than the physical distance and that is OK. From there I agree you want to time align the closer subs to the further one looking for optimal results. Optimal results means flattest frequency response but a users ability to determine that is really dependent on the tools they have available. Do you have SubEQHT, Minidsp or similar, SPL meter, Calibrated mic and software (REW omnimic, etc)? Are you OC? Do you enjoy tweaking and measuring?
I've commonly seen the following technique recommended, have done this myself, and it sounded pretty good. Play I sine wave at the crossover frequency (commonly 80 Hz) and tweak the distance/phase/delay settings to reach the highest SPL. You can even improve on audyssey's results so run audyssey for the furthest, tweak that sub as described, then add the next closet sub and tweak it's phase (or delay if you have minidsp or similar) for highest SPL, continue for additional subs.
Markus767 points out that with as much room interaction which occurs at these lower frequencies you really want to optimize the entire crossover region not just that single freq (the crossovers not a cliff like drop off but a roll off). Do some trial and error and make frequency response graphs (REW, omnimic, etc). Examine the 10, 20, or 30 Hz surrounding your crossover frequency and choose the setting combination that produces the best results. I'm very interested to find out if this is commonly producing different (which should be better) results than the sine wave at crossover frequency approach. I think when my 2nd FV15HP gets here I am going to play around with this approach and try comparing it to the sine wave approach.
Now If you don't have any tools you can do the sine wave at crossover freq thing with a SPL smartphone app. Don't trust the actual SPL values, just shoot for the higher ones. Clearly this approach is not for the OC's. Also make sure you do a sanity test playing a sine wave raising and lower the main volume to make sure the values are really moving around properly.