Markus and Andy, you should read this paper if you haven't already: Fazenda et al., "Subjective Preference of Modal Control Methods in Listening Rooms", J. of the Audio Engineering Society 2012, S.338 http://www.aes.org/e-lib/browse.cfm?elib=16324
. They did listening tests to evaluate the preference of different subwoofer placement and integration strategies, and source/sink (with a two sub setup) came out top.
Unfortunately I can say now that it's unlikely I'll have enough time to give you all the information I'm sure you will ask for. I have been rarely posting on the forums or writing my blog due to work and life being so busy.
The key benefit of source / sink over other approaches in rectangular rooms of consistent construction is independence from length mode related frequency response variations. Essentially with source / sink you don't have any length modes in the frequency range the sub array is operating. In a multi-row home theater (the main thing I am interested in, since these comprise the majority of the projects we do) this is extremely valuable, as all the other methods for integrating the rear sub(s) will result in some length mode variation. For example with zero delay on the rear sub(s) per Welti you will cancel odd order modes (1,3) but accentuating even order (2,4) modes. If you add some time delay to the rear sub(s) such as 2, 4 or 8ms or some other number then you can shift where the peaks and nips are (nodes and antinodes), and hopefully arrange them in such a way that there is more consistency in the response across each row. In larger rooms, where you have flexibility with row placement, then these latter two approaches can work well. However I've rarely found in practice that there is any flexibility with where the rows go. They go where they go because of other design elements: row-to-row distance requirements for reclining theater seats; distance from back row to wall behind (to get distance from ears to rear speakers); distance from front row to screen; depth of screen/baffle wall.
What you do have with a two sub setup (one front / one rear) is height and width mode variation, as you would expect. So you have height modes and you have width modes. If you have only one listener at one height then the two sub setup can actually work very well, as the Fazenda paper shows. I have data too I can post. Of course you would expect the response at seats which are along the width axis of the room to have different response, particularly if they the location of those seats starts to approach the null points for the 2nd width mode, because there is no width mode control in a two sub front / rear setup with the subs at midwall. If the room is wide enough, and the other seats on the row do not get near the null points for the 2nd width, then you can actually have very good consistency across the row. With a two sub setup (one front / one rear) you can of course move the front and/or rear sub away from the centerline of the room. This will result in some width mode cancellation but will make the response variation between seats on the same row worse (as you would expect for any non-symmetrical sub setup).
With four subs (two front / two back) at floor level the width mode variation is reduced by cancellation of the odd order modes. You can either place them in the corners or at the 25/75% of room width locations, with the expected typical differences between those two setups.
If you then take those four subs and move them to half height in the room you reduce interaction with problematic 1st axial height mode (unless your room is very tall higher order modes are not generally an issue). This approach is normally called CABS in the literature, though of course there seems to be some differences in use of the terminology, with one of the key researchers calling an eight sub setup as per the below CABS as well.
If you have eight subs, four front / four rear, then you have a DBA, and you get height mode cancellation as well.