I just haven't seen any good arguments not to. The best argument is that it would make mixed formations weak. But I'd argue that mixed formations are justifiably sub-optimal in most cases and the game should encourage you to separate tanks from infantry just like we do IRL.
Except that IRL, we don't, at least not universally? Example off the top of my head: until 2016 the US Army's Armored Brigade Combat Teams were built around Combined Arms Battalions which each contain 2x Mechanized Infantry Companies and 2x Tank Companies, and since 2016 I believe they've switched over to 2x Armored Battalions and 1x Mechanized Infantry Battalions but both of these use a mix of armor and infantry companies. The Stryker and Infantry BCTs don't use tanks largely for logistical, not tactical, reasons, and even then the Stryker BCTs still use light armored vehicles as part of the core of the formation.
More generally, combined arms has been known as the optimal force composition for large-scale battles since the early stages of WW2, although different powers figured this out at different points in the war. Certainly at the division level as a rule, and in many cases smaller combined-arms formations were used often on an ad-hoc basis (e.g. kampfgruppen).
That said, in Aurora it doesn't matter very much if you use combined arms battalions or single-type battalions organized into larger combined arms formations (divisions etc.), aside from some relatively technical points regarding breakthroughs and other such mechanics. Regardless of how you compose your formations at the lowest level, the targeting is essentially twice-random (random targeting of a formation, then random targeting of elements in that formation, both weighted by relative size/tonnage). This works very nicely from an RP perspective because in principle you can design combined arms or monolithic formations and either will work. If we make the element-level targeting deterministic, there's a clear metagame shift where monolithic formations become clearly optimal, assuming the formation-level targeting remains random (and if it didn't, we'd be opening a whole other can of worms...).
I admit I don't get what you mean by "weighting needs to be "continuum" in nature as a simple Infantry/Armor dichotomy does not exist in C# Aurora" since units are separated by whether they are infantry or light/medium/heavy vehicles already, making it easy for Steve to bias AT weapons to shoot non-infantry. I think there is more to this statement tho, could you clarify?
Basically, there are ranges of both element armor values and weapon AP values which have to be considered. It's easy to say "bias CAP to shoot INF, and bias MAV to shoot VEH". On one hand, what if the enemy force has LVH? Do we bias the MAV to shoot at VEH in particular, to get the best tonnage-destroyed ratio? Or just shoot at anything that isn't INF? What if the enemy force has a mix of normal and power armor INF, do we bias the CAP to shoot the weaker INF?
On the other hand, what about other weapon types? Autocannons and bombardment weapons are effective against light armor but not heavy armor, how do we want to bias these? More generally, how should we account for non-equal tech levels, which we would discover through battlefield intelligence? Do we adjust the bias of our MAV if it turns out the enemy VEH armor is greater than our attack stats, to focus on LVH instead? What about infantry LAV shooting at heavy tanks instead of static CAP nests, do we shift the bias there? Many more cases exist which could be considered.
You end up at a point where we either need to write a complex set of bias rules for every weapon (which are likely to be controversial, to boot), or else bias the targeting to be basically perfect by having every weapon target the element it can score the highest tonnage-weighted kill rate on. That's not to say that the latter is automatically a Bad Thing™, but it would substantially rewrite the balance of Ground Forces (meaning a lot of balancing work for Steve) and in my opinion at least would eliminate a lot of the flexibility we have right now in designing reasonably effective formations.
Ultimately, the idea of "anti-infantry weapons shoot infantry, anti-tank weapons shoot tanks" only works if you have two kinds of weapons and two kinds of units. Aurora does not have this, we have six basic unit types (most of which can vary in armor level as well) and the same number of basic combat weapon types with varying sizes. That's what I mean by "continuum" weighting, we have to come up with a system that works for all weapon and unit base types without leaving easy exploits or suboptimal behavior. The nice thing about random targeting is that you won't
not target the thing you want to shoot at, you just can't target it
reliably. That's not exactly unrealistic, particularly given the rather abstract, high-level nature of ground combat in Aurora which precludes a lot of the tactical detail that actually decides what your guns can shoot at in a real battle.