It's not necessary to use a VCA before a filter. However, if you have one that can be placed there, you can then do things such as routing a single VCO through it and then into the VCO mix, and by giving that a longer attack than the post-VCF one, that'll let you add harmonics/VCO density/detunings/etc as a ramp-up as long as the note-on is present. Or even more things; that's just one example. VCAs are the unsung heroes of modular synthesis, really...they allow all sorts of trickery involving controlling audio levels, CV voltages, and the like that wind up upping the complexity of your patch. But it doesn't convert the VCO signal to audio; the VCO outputs audio itself, and the VCA (plus a modulating source) can control that audio level. Also, keep in mind that the six channels of the HN mixer also have VCAs for audio level control
The Synthacon VCF looks simple...but, having owned a real Steiner-Parker Synthacon for many years (traded it for an ARP 2600 quite some time ago), I can tell you that it's anything but, sonically. It's very versatile, does everything from punchy bass to rip-your-ears-off screeches, and pretty much anything in between. There are definite reasons why the Arturia Brutes use them; I put that in in there so that you have some sonic consistency, but with the added potential of two different simultaneous options with the same filter architecture.
The Klavis VCOs are rather deep, functionwise. But I wouldn't toss the Batumi out just because the Klavis oscillators can tune down into subaudio. Batumis can do tricks with their LFO signals (especially with the Poti expander) that lots of other oscillators can't, such as operating in a quadrature mode or in other phase-relational modes on a single frequency. They also have an internal divider mode that has lots of potential for clocking applications. And yes, you can do a lot with the Klavis VCOs + the Shifty, even beyond a 4-stage shift register. For instance, you can switch the Shifty into 2-stage, then have two VCOs per outputted CV for two-part duophony from the same initial incoming CV. And again, that's simply one possibility among numerous ones. Or as you noted, drop all four VCOs on the same pitch for big detuned punchiness. Or three, then have a single one on top as a lead. The list goes on, as you can see.
As for the Zadar...it's nice, but the fact is that, in the early stages of using modular, you're going to want to have modules that require lots of tweak attention and manual manipulation potential. I've always felt that it was more important to reach out and get a handful of parameter control rather than futz around with menus and assignable controls. And you will definitely find uses for the A-140-2 ADSRs (especially given their jumperable custom configuration possibilities) all over the place, even in tandem with the Batumi's LFOs. You could even use the latter to control the time aspects on the former, varying envelope lengths in cyclical fashion. But for the synthesis parts in that build, I went with keeping everything as 'hands-on' as possible, with the only menu-driven device being the Gatestorm sequencer for the percussion sounds, because sequencers need more direct visual feedback to the user than in most other modules, and also because it's very useful to have recall for your sequence loops and you need to see what's being recalled quickly when switching sequences while playing.
Anyway, this is an easier to learn build, as you've kinda noticed. It's possible to get really deep into sound design with this, but all of the controls remain grabable for the most part, making it also more adept for live work in addition to studio. And it's also a build that's more open-ended, allowing for further augmentation in the future with more cabs, devices, etc. Putting it together with the Mini 2S will result in a very potent sequencer-based rig indeed.