The notion that you need 30 samples comes from the fact that if you are doing the basic Z test (the one they start you out on in basic stats), you need about 30 samples to meet the assumption that you really know the long term mean of the process. This is also about the point where the T Test (which everybody uses) and the Z test (which practically nobody uses) merge and become the same. It has nothing to do with the sample size needed to estimate ES or SD.
Measures of dispersion, like range (which shooters call ES) and standard deviation are harder to pin down than estimates of means. For example, I just ran a simulation with 30 data with SD = 15, and the estimate of the mean was only reproducible to between about 11 and 18.5. You can get a good estimate of the mean with a smaller sample than an estimate of SD.
SD and ES can be converted from one to the other. For a sample of 5, divide your estimated ES by 2.3 to get an estimate of SD.
SD of MV usually matters very little, so long as it is something reasonable, say below about 35 FPS. Since variation does not add linearly, the total variation will be almost entirely governed by the largest source of variation, which is normally not MV.
Basically, muzzle velocity variation only matters if the rifle and shooter are able to shoot the normal distribution (precision) at a range which velocity variation results in more drop than the rifle's level of precision.
If a rifle is a 1 MOA rifle, velocity variation won't be critical until the standard deviation could result in more than 1 MOA of drop variation.
