As far as classes go and inasmuch as it pertains to power/heat dynamics without going off-topic, you can think of it like a square/cube tradeoff.
Reactor classes increase geometrically, hypothetically in nice clean cubes: 1^3 (1c); 2^3 (8c); 3^3 (27c); 4^3 (64c); 5^3 (125c). We'll assume class ratings of components scales from 1 to 5 for ease and simplicity. While the size increases only slightly between each reactor, the comparable volume and thusly the comparable power output (plus efficiencies from using a larger single package) tail off as class rating increases. Principally, I'd make the prediction that you're more likely to see redundant reactors on larger vessels where the redundancy itself (and possibly cost reductions) outweighs increasingly minimal differences in size and output between the larger reactors and you're more likely to see overlarge single reactors in the smaller ranges of craft most notably in glass cannons where volume will be at an assumed premium.
With a radiator two of the dimensions with increase faster than the third and I don't really expect the things to get larger than 1c thick in the first place. Correspondingly we will see the oft-elaborated need to maximise surface area in order to take advantage of greater heat dissipation needs in larger and more powerful vessels.
Weapons are even further distorted towards linear power growth over class rating increases, principally you can make the round longer than it becomes wider to take advantage of penetration effects, hitbox profiles for missiles, and fuel economy (again for missiles). So while linearly compact, firepower might experience a somewhat more geometric increase over class rating. However while any level of "alphastrike" firepower is attainable fairly cheaply, rate of fire in cannon, velocity/maneuver in missiles, and beam concentration (i.e. range) in energy weapons might suggest geometric cost-curves for a given level of performance. A cheap solution to the performance curve is saturating your platform with weapons emplacements and relying on numerous but inexpensive attacks at closer engagement zones, but that becomes its own cost-curve for given levels of performance at the end of the day as well.
All other systems except beside senscomms, thrusters, and perhaps large-scale life support seem like they should have a near linear use-to-volume ratio. How are we even doing gravity? Will gravity wheel systems work as a specially coded scalable prefab? Will the falling mechanics and induced gravity be a function of acceleration/velocity? Or are we using space opera gravity and building wheels just for show?
Reactor Classes: cubic size disparity, high cost to increase performance
*Thrusters: most are a reactor, or a series of reactors, with one end open to space, otherwise near-linear tankage-to-thrust input until we start using Newtonian fuel accounting
*Life Support: assumedly volumetric performance ratings for "atmo-generators"
Radiator Classes: squared size disparity, linear cost over surface area to performance
*Senscomm: they're a type of radiator with a bunch of computers attached to the back end
*Life Support: filtration equipment relies on membranes similar in function to radiation setups
Weapon Classes: near-linear size disparity, low cost to increase basic firepower per unit of munitions (i.e. "per kg of bullet"), high cost to increase secondary firepower factors (velocity, cyclic rate, maneuver, dissipation, etc)
P.S. Getting back to whether or not recoil would be a decent addition or an unnecessary complication, my personal opinion is that it adds a special depth to designs that doesn't just rely on what sort of gear you're kitted out with, but it ties into design necessities in the same way as you would plan your thruster placement. Even if compensating devices are available to install, a well-designed ship that doesn't need to rely on those extras will be that much more efficient and gives builders a special niche to focus on within the general grind to the game's progress. Even further it opens up the game to quirky designs that, by choice or accident, become known for unique and possibly even desirable maneuvering profiles.