Hi @JulianG
Just chiming in: you probably already know and have tested it, but @JMPicard idea of having a fixed “stencil” (he was suggesting a sunflower shape) rather than actual Monte Carlo simulation is a good one, to the best of my judgement.
You could weight each (fixed) point by its probability value, because you have access to the probability density/mass function: that would give you a much better estimate than using Monte Carlo simulation (assuming the same number of samples/points), and it would take into account more accurately the effect of cover.
Bursts can be taken into account by just multiplying the weight of each point (the probability) by the number of bullets: that would allow to compute tje expected value of the damage for each stencil point. The nice part is that with this trick you need to check for occlusions (I suspect this is the most expensive step) only once per stencil point. With a little bit of extra math effort it may even be possible to compute quantiles (and hence a damage range).
Just my two pence, though.
Best regards (and great work so far!),
Michele
P.s. a while ago I tried to reverse-engineer your sampling strategy in this post. In this reply I am assuming that you are doing something similar, which would allow you to pre-compute the probability of each point in the stencil before a mission starts. I am also assuming there is no mechanic that changes the trajectory probability between bullets in the same burst (e.g. no recoil)