When in college I used one of those little cherry smoke bombs to provide thrust for a paper airplane. It climbed and flew along like a little jet. Then caught fire. Not the safest experiment!
With the smoke bomb, that little hole is actually too large to cause useful nozzle effects. You're just getting mass flow and a little pressure. Therefore enlarging it will only alter the burn rate a bit perhaps, but not really alter the total impulse.
It takes very little mass imbalance, drag imbalance, or thrust imbalance, to cause a small short light rocket to deviate rather quickly from vertical. You need all the forces and the center of mass to be colinear.
Gerald
A smaller hole causes the pressure to be greater (for a given burn rate) and therefore the exit velocity as well. That's why putting you thumb over the end of a hose and allowing a small exit gets you much better range on the water "jet". And the smaller jet hit's your target with greater force, which indicates (Newton's third law) that its thrust is grater as well. Of course the simple hole is not a compression-expansion nozzle (we are in agreement) so none of the subtleties of a good nozzle apply.
The exit hole on those smoke bombs is (if the image in my head is accurate; it's been a long time) about four mm. The ID of an 18 mm casing is 13 mm. That's roughly a 10.5:1 area ratio.
Sounds like a ground test is in order. Make a smoke bomb (Abby Hoffman's recipe should do nicely, or a slow burning BP such as with baking soda filler) with an 18 mm casing plugged at one end, and put it on a test stand.
As for off-center drag, there's no reason to expect any if the pods are made well. And, launch lugs make little or no difference (granted, they are much closer to the center line). Poorly build rockets with things like random slight canting of fins and other defects, up to a point, still fly safely if not well.
Finally, with regard to the mass imbalance, remember that we're not talking about the mass of a smoke generator in one pod and not the others. We're only talking about the transient imbalance due to different burn rates while all three are burning. The burn rates should be essentially the same assuming they are all the same composition, so the mass differences will be trivial. Come to think of it, the same goes for what little thrust there is; it's not one pod's thrust that matters but rather their transient differences; the differences between small and similar quantities is is tiny.