Glad to hear!!! If it helps, I can start a spreadsheet of parameter IDs I tested and the results. I used the form:
F0 18 0F dd 55 01 02 xx xx yy yy F7
I used dd = 7F.
For preset select I sent -1 to parameter 897 to use the edit buffer.
For layer select I used 0 for layer 1 (parameter 898).
Tested:
Next-Note change: Layer Instrument, LFO 1 Shape, Layer Cord Amount [I used DC -> pitch because its easier to hear!], Filter Type, Layer Pan, Filter Cutoff, Filter Q, and maybe some others I didn't write down.
Immediate change:
FXA algorithm [both master and preset], FXA Decay.
With direct parameter changes my guesses are that:
--most if not all preset layer parameters will not be immediate,
--some of the preset common parameters will be immediate
--most if not all of the FX parameters will be immediate
I am pretty sure I tested the 4 FX send percentages, but I didn't write down the result.
I did not test arp pattern changes, but assume those are next-note effects.
I'd be interested in the results of any results you get.
BTW, to test I used an audition riff at BPM = 1 and sent the individual sysex banks from SONAR.
As for embedded sysex, I can't imagine them not working. Hmmmmm. Maybe the next test will be a drum pattern using a 4 layer preset with track 16 as a ROM Instrument Control Track with embedded sysex to vary the 4 layers. Hmmmmm.