These look damn good.

I need to give you my old UE-10 so you can measure that. The UE-10 was the precursor to the UERM and at the time (2005), it was Jerry Harvey's top tier CIEM.

So you say that if I take FK and WBFK separately, they will be way better? I smell opportunity...
Or if I changed the cross on there to use them as a true dual, would the peaks disappear, as the fields become coherent?

Your best bet is to use a DWFK (yes, a dual WBFK exists) for upper-end bandwidth. However, the driver usually has high-impedance SKUs, and the peaks will still exist, you just have more headroom to tune them down. You can then use something else for the midrange.

Interesting about 1723. It seems like what I was considering doing, but I thought to use CI-30050 as the bass (throttle to 1.5k? what's the harmonic distortion there?); 26A007/9 as mids+highs (best THD, great FR) and 2354 as the tweeter (best extension), maybe x2. All low impedance meaning more headroom for impedance and crossover tweaks. Alternative bass driver is FED-30048-I04 - thoughts on this one? Its FR is not as good - I could use that 1k resonance of CI. Any ideas where I can get small numbers of the Sonion BAs for experiments? Otherwise, I'd have to skip 2354(s) (real shame) for WBFK, replace 26A007/9 with inferior 26A005/8.

AFAIK, the FerroFluid drivers are not good for audio. The FerroFluid really is only good for shock protection.

Why use a 2354 over the 2389? The 2354 is just the vented version of the 2389; on an IEC60711, they'll measure identically above 1k, basically. In fact, because of the venting, the 2354 will probably have a little more THD in the highs too.

I have not had the opportunity to try a CIEM.  How do the UE-4's compare to something like Luis' Paradox?

The UE10/JH10x3 use different drivers from the UERM.

The UERM uses two Sonion 2389-type high and mid drivers, with different high-pass cap values (to space out the mids and highs correctly), and a Knowles CI driver (similar to the CI-22955, but not identical).

AFAIK, the UE10 uses two Sonion 2000-type (similar to the Knowles CI) drivers and a Sonion 2389-type driver for the highs. This is a similar driver selection to the UE7, except that the UE7 is two-way, meaning that the two Sonion 2000-type drivers are twinned (in parallel or series), while the UE10 is three-way, so there's a high-pass applied to the second Sonion 2000-type driver for the midrange.

The Sonion 2389-type high driver in the UE10 likely also has a larger cap value, e.g. 2.2 uF vs 1.5 uF, compared to the UE7, as the UE7 is super sparkly and kinda mid-recessed. From all indications, the UE10 shouldn't be this way. We can also predict that the UE10 will have a warmer, thicker midrange, as is customary for the Sonion 2000 / Knowles CI sound (CSD measurements may be more telling in this respect).

If the UERM was derived from the general sound of the UE10, then it was that they probably started off giving the Capitol guys a sample hearing of the UE10, and went from there to tweak the sound signature of the UE10, eventually completely replacing at least a couple drivers and reworking the entire crossover network.