Also, another issue this size progression will bring up is that it makes 'canister' launchers (ie. a missile containing other missiles as second stage, and no actual range/etc on the first stage) very attractive. If a Size 9 launcher is 3 HS and a Size 3 launcher is 1.73 HS, then it becomes extremely attractive to design size 9 'canister' missiles containing 3 size 2.95 (or whatever the exact size ends up being) missiles inside, getting vastly higher salvo density (~6-6.5x) and ultimately doing nothing to actually change the small vs large missiles issue, just shifting the meta for how it is done.
Yes, that is a very good point, especially if you start designing really large missiles. Size 36 is only 6 HS - 2x larger than size 9 - and you get 4x more sub-munitions.
EDIT - I don't see an easy way around the above, so what is needed is some capability that requires internal space and is a reasonable alternative to a larger missile wave. Onboard ECM/ECCM is already an option. Perhaps other warhead types could be added. Laser heads that attack from a specified range depending on warheads size and laser tech, or shaped charge warheads that are larger than normal but with improved penetration, or Tandem-charge for a similar effect. Maybe missiles with retargeting capability if they miss, or some form of evasion capabilities, or missiles with electronic damage similar to microwaves (that require large warheads). Perhaps it's time to revisit EW and add jamming and counter-jamming. Open to ideas.
Just scale the size of the "submunitions" component of a MIRVed missile to the square of the number of submunitions instead of linearly. That way you can fire 4 size 4 missiles from 4 HS 2 launchers (8 HS total) or from 1 size 64 canister fired from a single 8 HS launcher.
The in-universe justification would be that TN warheads and engines don't play nice with each other when they activate, so you can't stack them like sardines in your MIRV components.
Note that the above scaling is so punishing that it completely negates "canister" MIRVs for alpha strike purposes, but at the cost of making MIRVed warheads (canister or not) strictly inferior in terms of magazine space: A size 21 missile with a size 5 bus could carry only 2 size 4 terminal stages, which is fine in terms of launcher size (the two size 4 terminal stages launched as two separate 2-stage missiles with a size 2.5 bus each would cost just over 5 HS of launcher as opposed to just over 4½ HS for the MIRV), but take up 21 MSP worth of magazine space as opposed to 13 for the separate missiles.
If instead we scale to the power of 3/2, the same 2x4 MIRV on a size 5 bus would yield a size 16½ missile vs. 13 for the separate two-stage missiles (but the bus would be more efficient due to larger engine size), and require 4 HS of launcher as opposed to the 5 HS of launcher for the two separate missiles. That trade-off between launcher space and magazine space does not seem unreasonable to me.
Taking the 3/2 power scaling to the extreme of a size 64 canister of size 4 missiles, this gives us 6 size 4 missiles for 8 HS worth of launcher, as opposed to the same 6 missiles taking 12 HS of launcher to launch separately. This seems still a bit exploit-y, but we can combine it with a minimum bus size (again, TN warheads do not play nice with each other in MIRVs; need scaffolding). If we impose a minimum bus size of 25 % of the total size of all submunitions, including scaling penalty (a minimum that will not be constraining for reasonable MIRV designs), then a 6 size 4 missile canister grows to size 80, or 9 HS of launcher as opposed to the 12 HS of separately launched missiles.
If 9 HS vs. 12 HS of launcher looks exploity, recall that we have now well over
tripled the tonnage of colliers needed to keep this vessel in supply for a repeat performance, in order to get a 25 % reduction in warship tonnage. Now, magazines are cheaper than launchers and colliers are cheaper than warships... but probably not by a factor of 6 (assuming roughly 1:2 collier:missile warship tonnage as baseline).
So what this dynamic gives you is that huge canister MIRVs makes sense for box launcher vessels operating out of populated systems that can produce their own resupply of munitions, but very quickly become impractical for sustained power projection, or even for establishing forward bases for patrol vessels.
That does not strike me as an entirely undesirable dynamic.