Hard to say exactly what that artifact could have been. It could have been an interference signal that was present during the original recording. Both the stock and modified opamps would have sufficiently high PSRR, and the caps added don't do anything about power supply noise or interference. So either the source of the artifact is not related to the mod, or there was likely another change involved in the mod.
The mention of input transistors could change the noise analysis materially. Discrete transistors are frequently used as a first-stage in front of an opamp for improved noise performance. It takes more space though, so you might not see it as often in portable gear. But it's a cheaper approach than paying for an expensive low-noise opamp (and solves a related problem of input topology). Since there is gain in the first stage, the noise of the opamp becomes relatively lesser important as first-stage gain increases.
So if the 70D uses that type of input circuit, then opamp noise might not matter except at very low gain where you don't care anyway.
There is definitely nothing else done beyond swapping all of the opamps and adding the poly caps. I took the instructions from Jim, bought the parts myself and sent them off to the SMT shop. Again, the full details with a parts list are on the FAQ page for those who haven't seen this yet.
Investigating further into this spike I'm seeing: Most, but interestingly not
all pre-mod recordings made with the 70D at 48kHz sampling rate have a spike at 15,734 Hz, with the left channel peaking about 10-12 dB higher than the right. The levels of those peaks I'm seeing are around -63 to -74 dB for the left; -73 to -83 for the right. This is with HIGH gain setting, typically with the gain 1/3 to 2/3 up.
96 kHz recordings (the very few I have) move that spike up to 21,903 Hz or 21,904 Hz. I had done a test at this sampling rate recording with my CM3s at max gain for each of the 4 levels, just sitting open on the carpet in a quiet room. The peak reached as high as -63.5 dB at the HIGH+ gain level, with each of the lower 3 gain levels proportionally lower according to their maximum gain setting.
This pops up in a variety of recording venues. It's also not the mics, as I analyzed a couple 4-channel recordings that also used the X-Qs and I got the same exact thing, just higher in level.
All recordings I have made with the 70D post-mod do not have this spike at either sample rate.
However, I did find a very small number of recordings with my FP24 > M10 chain also (again using different mics), suggesting either there is something going on with either of those devices which similar to what I'm seeing in the pre-mod 70D, or it's somehow random and environmental as you suggest.
I can't find anything interesting about those two frequencies above 21kHz, but 15,734 Hz is the NTSC horizontal refresh rate. It would make sense to get that spike if I was ripping the sound from old TV shows...