.. my question is really about how it sounds, given that the Side figure-8 will be severely off-axis. When I do "normal" Blumlein, each cap is 45 degrees off-axis for sound coming from dead center, this method will have the front facing figure-8 directly on axis for such sound. If I were recording just two instruments, and the "normal" Blumlein figure-8s were pointed at each one so that one is in the left channel and one in the right, they would each be on-axis; with this setup each instrument will be 45 degrees off-axis for each cap. I can't believe it sounds exactly the same..
I've wondered this as well. Mathematically, Blumlein and M/S with two fig8's is equivalent. You could send a Blumlein recording though a M/S encode/decode chain and adjust width just like the M/S recording.
This would be interesting and simple to test. Tape one set in Blumlein, rotate rig 45deg clockwise at set-break, Tape second set M/S. Matrix the second set with equal weighting and compare. If the mics are set up truly coincident, the only difference is the quality of your summing when setting up the matrix (pretty much a non-issue when done digitally), the and the on/off axis observation you mention.
Most multi-pattern mics have more accurate off-axis response in fig-8 than other directional patterns, so the 'on/off-axis' difference you'd be comparing in the test may not be as significant as comparing a more typical M/S recording using a cardioid mid (dematrixed with equal 50% weighting) with the equivalent X/Y configuration of hyper/supercardioid-ish patterns set at a 120 degree recording angle. But there is
no way to really set up that comparison accurately.