Just read through this and it seems interesting.
One question I have is whether separating this ARL++/ZML thing from Quark itself is worth the engineering effort.
I believe it's quite rare that modders will use another modder's core library instead of making their own (PAUCAL, ARL, CoFHCore, etc....). I myself tend to avoid all external dependencies as much as possible in Patchouli/Botania to avoid getting broken when things change. Hence, the scope of Zeta's reuse would be limited to other VM mods. Patchouli/Botania/Neat/Psi don't use ARL and I see no reason for them to adopt its successor; the other small mods like Akashic Tome or Morph-o-Tool use ARL purely based on inertia and would probably actually get cleaner if the ARL parts were removed.
The only case I would see the scope of reuse of this new system being significant is if Violet Moon created a new, significantly large, crossplatform mod that would also need this system besides Quark. I don't think there's plans for that atm?
Another point to consider is that introducing external customers to something means you now have to care about them when making changes of any sort. In the past, you could refactor the module system and update Quark in tandem in one atomic commit, but if you have external customers you have to go through a lengthy deprecation cycle at least one major Minecraft version in advance. It will significantly slow evolution of the system.
So I think the plan for refactoring Quark to be multiloader and cleanly separating the module system from the loader is a necessary step, but actually shipping that as a separate thing outside of Quark can be a significant commitment with unclear benefits.