Agreed on modularity being a feature of the future. Logistics wins wars, and simplifying your logistics with a singular model versus several dedicated, if capable enough, can yield enormous gains.
I'd note that even today's latest and greatest can't field everything though, F-35's currently have no HARM capability, starting with the point that the missile won't fit its internal bay - given the objective of stealth and the penalties of carrying externally, its not wired to even try. AIM-120's have external mounting capabilities though, no reason to hobble it in lower threat environments. Even as modular as they are, weaponry included, the ever present issue of aerodynamic and torsional stresses, wiring/connectivity and internal software/awareness of the munitions envelop remain a thing, though we certainly do try harder to keep it an option. Fortunately for the concept, this is the sort of detail aurora abstracts away entirely, absolving us of needing to dwell on it, once sure we don't consider it too large a hurdle.
I also agree that it may be difficult to balance well, and especially the AI using it effectively. Though an AI effectively abusing this could have very capable carriers. An entire fighter wing that can launch with the right armament to best make you suffer, and then be completely different for its next engagement?
I would contend it should not be possible to internally mount a complete weapon system, and still be able to later swap it out like a munition - once inside the armor, that's that. Something mounted outside the main hull may be open to a little more variability, but you lose the advantages full integration would have allowed. This tradeoff would naturally be in one of two forms for gameplay reasons, either a loss of performance for same mass, or a growth in mass for same performance. Given the intent to mount these on fighters, the loss of performance for same mass seems the better candidate. So for the same mass, you get a somewhat inferior, but more flexible weapon system.
In the interests in achieving reasonable fire control range, the fire control should not be in the mount, it should be hosted by the parent craft.
Now, on to energy supply. Under the assumption that modular mounting of lasers or particle weapons is a thing, should the power be in its own pod, part of the external mount, or mounted in the parent craft? I'm partial to requiring they be internally mounted, and suffer a penalty to delivery efficiency, perhaps a simple percentage loss of power at the mount, requiring a larger generator for same system, or the more explosive flavour for similar/same size.
This leaves you deficient compared to a dedicated system in a couple ways. One, the weapon system will not have the advantage of armor - not that fighters have much anyways. Two, it will lack the performance of a dedicated system. Three, even when not mounted, you have mass spent on the no unnecessary power generators. Four, the expense of design and construction should be larger, its one thing to integrate the weapon, its another to run it from a mounting, the design considerations change.
In an all up fight, the fighter with permanent mounts should perform better and cost less - it sacrificed flexibility to get those advantages. There are many efficiencies you can leverage when you can optimise for a specific system rather than generalising.
There is one serious caveat to this line of thought, and QuakeIV already mentioned it. If it is reasonable to mount a weapons system, why would it not be reasonable to do something less challenging: mount active or passive sensors, ECM, or ECCM? What about even simpler things like fuel tanks? Could fighters end up serving as emergency tugs dragging out a podded tractor beam?
Personally, I think its much harder to justify the can of worms for full weapon systems, much less difficult for subdivided box launchers.