I use M/S a lot, not just the main stereo pair. I would agree with Ealing if the Middle mic is an omni, because the side lobes of the AEA ribbon are approximate spheres, so as long as another mic collects sound from the front portion of the spheres, all will be well with the algebra (Mid + Side = (L+R) + (L-R) = 2L) and (Mid - Side = (L+R) - (L-R) = 2R)
However, if the Mid cardioid mic, whose lobes are much less spherical, is pointed along the longitudinal axis of the AEA "like two fingers" then I would expect the stereo image to be less successful.
I have found using an omni, km83 or gefell m294, in a dry smallish room preferable; in a stoney church, a cardiod gefell m300, has yielded better definition. I track all mics straight to recorder and un-matrix later in DAW but I monitor the split signal (usually unbalanced 2nd output from pre-amp) into a small mixer and use a phase reversed 12 inch patch for the side -ve signal;
A useful tip is to turn the Mid off, pan center the Side + and - signals and adjust the gain until you hear nothing: note the relative positions of the faders in case you have to go up or down. Now pan left & right for width: bring up the Mid until you have a pleasing stereo image in the L/R buss. Don't use the mixer to adjust the gains, use the outboard pres and you shouldn't have to re-adjust the side signals. If you can route the three signals to one Aux buss, then you can control the whole thing w/one fader.
If you want to add another M/S pair, I've found it usually better to mix the side signals to taste first and then the Mids; then de-matrix the whole thing once. If you de-matrix to L & R first you'll run out of inputs & busses. You eventually learn to judge how a M/S signal will "pan out" just by judging the reverb-to-direct ratio but a $50 monitor-only mixer for each pair works & is a lot cheaper than a $500 decoder.
WT