Right...one key, I've found over many years, to making a synth behave like an instrument is that if specific aspects are grouped by their particular specialty (generator, modifier, controller, processor), then the flow around the instrument becomes clearer and more intuitive. This, in turn, changes up how you approach the instrument; instead of having function 'A' all over here and 'B' down that side or such, mingle everything together as far as playing function, but make it cohesive as far as actual functionality.
Another thing that will help here a lot is fixing the output stage. Right now, there's a couple of manual mono mixers, one of Intellijel's new sorta-stereo ones, and the output isolator itself. I really strongly suggest making the output some sort of stereo performance mixer, something that also allows VCA control over the strip levels at the very least, as well as panning if you can swing it. Shift the mono mixers to use as submixing into the final stereo strips. This way, the mixer becomes something of a playing surface, allowing you to better slot various instrumental signal paths within the cab into a cohesive stereo-out mix.
As for the AJH VCOs...one thing that will click that sound in even tighter is to get some sort of Moog CP3 clone. The CP3 circuit, if cloned properly, is actually not 100% clean; it introduces subtle nonlinearities into the outputted result that actually emphasizes certain euphonic partials. It's not distortion per se...and a bit difficult to describe unless you've ever gotten your hands on a Moog device with that particular mixer circuit (such as a modular, or the Minimoog).
Now that I've got a clearer idea of what's actually going on here, let's see...
[Had this all typed up, and MG logged me out in the time it too to write it up. But I have my own word-processor, so...take THAT, rotten, evil login script! Ha!]
OK...you will notice that there's definitely some things missing from your original build. I did tinker with this a good bit, but I kept your primary functionalities as you'd posted them pretty much intact, even improving on the ergonomics of them. The entire order of this got a lot of reworking, too...but it now has that 'subsystemic' flow that I think you'll find makes it all easier to program and play, especially with the module swaps.
Row 1 has your buffered mult and adder/mult to the far left, with the AJH glide/noise module there so that your can easily impose slewing on CVs as needed. Also, with that positioning, you can easily place the slew limiter into a CV path, then send that back to a mult and distribute it that way, giving you a couple of nice options. The Minimoog VCOs were dropped to three (which is like the real thing) and the CP3 clone from Manhattan was added. This now replicates the VCO-mixer part of the Minimoog signal path...and there's more coming about that, so keep reading.
Digital-type sources are on the right end, then the Mixup was paired with the Quad VCA so that there's a lot of possible mixing/VCA control routing options present, given that the PDO and Shapeshifter have multiple outputs and there may be different ways you'll want that working for different situations. Oh...and a mult. You needed a passive one.
Row 2 starts with modulation sources: one Disting, your o&Cs, Batumi, Maths, Quadra. I then added three Ladik ADSRs with normal and inverted outputs, which eliminated the need for the attenuverters and added three desperately-needed 4-stage EGs. An Intellijel uFold is next for waveshaping/distortion, but with CV control over waveshaping in up to three stages. Then a SE 5089 VCF...and this is the rest of the tale on the Minimoog signal path.
That filter is a pretty effective clone of the 4-pole Moog transistor ladder lowpass VCF. If you feed this directly from the CP3, you then have the actual Minimoog audio signal path, up to the final VCA. Because of this, you get the nonlinearities and euphonic...something...that the Minimoog proper tends to have. Mind you, Minimoogs had a lot of variation to them because of component tolerances and such...but this is now a pretty spot-on redux of that audio path, hence the ability to remove the fourth VCO. The two new components (especially that 5089 VCF) will more than make up for its absence.
Your Polaris is next...and now, you have two excellent VCFs of two 'flavors'. Quad VCA next to mix the VCF outs...or you can just as easily send them through the LxD, which I relocated right beside it. Or both...? At any rate, the Quad VCAs were also set up this way so that you could both use them as exponentials for audio level control or in linear mode for CV levels, and split out your mixing duties but have one or two VCAs left for linear work.
Row 3: the other Disting, placed here to work more with the quantizing modules. Your drum/bass modules come next, then a stereo mixer to sum these down and get a good stereo placement so the Bitbox has a nice stereo image to screw around with. Note that all of this is now directly over the sequencing/control section for easy routing.
Effects processing is next: looping delay and Z-DSP. This closes out that row...but notice, again, the placement: these processors are directly above the mixing/output section, again to make routing easier. Plus it allows for some routing options with ease that we'll get at in a bit.
Row 4 is control. MIDI interface, then the Pamela's, Eloquencer, Varigate and Bow Tie. The idea here is that, with things laid out like this, you can easily send in a sync clock via MIDI or USB, use that as a master, then the next several modules get to tamper with that incoming clock as needed, all in a nice line so that if you want to do some screwy things with that clock, it's easy to do. And as mentioned, the placement of all of these is right below the quants, drum and bass, and Bitbox so that all of your functionality of these things in tandem are right there, simple to route and use as a unitized whole. The Bow Tie was placed down on the end of this so that you can now easily use it either for routing CVs or audio, whatever you feel like. Plus, anything that needs to go to the VCOs can now just flow right up the left side of the cab, keeping those patchcables out of the way. Much cleaner.
The Triple VCA is in place so that you can use it to VCA submix a single strip going into the Toppobrillo Stereomix. That thing there is going to make a huge difference. It allows CV control over audio levels, panning, and AUX sends on four inputs, which also have some routing tricks such as a cue send. This also has a single AUX send (mono) and return (mono or stereo) loop, so it's possible to parallel-process part of the overall mix through quite a few things.
Your MSCL is next, to put it in place to easily dynamic-process the overall stereo mix. But since the new Happy Nerding OUT has two stereo inputs, you can do some parallel processing work with that, or send in a separate signal path, or...well, a lot of things. That aux stereo feed has a pan control to adjust the stereo image, but it's fixed-level. The main input, however, is variable, so you actually get to use that OUT as a final stereo summer, in addition to being stereo balanced outs, a headphone (1/4”) amp, and output-level metering.
So, yes...I yanked several modules out, some of which you'd mentioned but a few of which I felt were not really well-implemented, or could be done better in a smaller space (pulling the Malekko distorter in favor of the uFold, for example). Overall, though, the idea behind what I'd cut was to maximize what you'd described as the 'mission' for the build, and I think that you'll find that this arrangement actually makes it more intuitive for those, as well as suggestive of some other possibilities not really clearly shown in the original layout. But the main split here is to get the 'synthy' bits up top, and the 'controlly' bits down and all together on the lower tiers.
And given what the 'pulls' might bring on the used market, you might be able to do this for close to zero-extra. Or maybe even a little left over...I did come in at $432 less than your figure, after all!