I'd like to suggest a consideration here, in general, for those designing rocket recovery systems. Many who consider this problem seem to do so from the perspective of everything working nominally. That is a dangerous assumption when it comes to rockets and rocket flight. Assume the anomalous, assess the probability, and then make the design decisions.
Suppose the not all that uncommon situation of a rocket which does not end up with a perfect trajectory relative to the air mass. That is, it has some horizontal component of velocity compared to the air. At apogee such a flight will achieve zero vertical speed but the horizontal speed will not be zero. The higher the rocket goes, the less stable the rocket, and the greater the wind at launch, the greater the EXPECTED horizontal velocity at apogee.
So, there are TWO mechanisms promoting premature deployment of main at apogee. The first, which is what has been considered a little bit in this thread, has to do with the relative masses of the nosecone (or connected forward section) and the motor section of the rocket, their relative velocities at ejection, dissipation of that energy, and the resultant shock transmitted to the retention mechanism for the main chute.
The second is the shock generated upon opening of the drogue at non-zero airspeed. This is dependent on the speed, the size of the drogue, the drogue's opening speed, the drogue's drag coefficient, etc.
So, an engineering approach would be comprised of two parts.
First, to decide on a maximum expected horizontal velocity component at apogee and then apply a safety factor. Simulation for nominal flight in max expected crosswind with a variety of motors and a few degrees for non-vertical launch rail can give the horizontal speed range at apogee. The largest of the simulation results is still in the EXPECTED range. So throw a safety factor in there. 1.5x? 2.0x? Decide. Now throw in a couple second late deployment, a second or two for drogue opening, do a bit of vector math, and find the velocity. From the speed and whatever info you have on the drogue, determine how much force the drogue can then generate.
From there, determine the number of appropriately sized shear pins it would take to not have a premature deployment.
Second, Determine the drag force max on the rocket trying to pull the nosecone out early for a premature drogue deployment at motor shutdown. Simulations can help there. For simplicity, assume the nosecone drag is zero and then apply a small safety factor for assurance. Determine the number of appropriately sized shear pins it would take to not have a premature deployment, for both sections.
For the pins retaining the main, you now have an answer.
For the pins retaining the drogue, take the maximum of these two answers.
Now determine how much ejection charge is required to reliably shear these pins for each section.
Etc.
Gerald
PS - Nothing in this approach handles anomalous situations. This is just to handle EXPECTED situations. If you want overkill, scale up from there. Overkill is not a bad plan. Someday you might need it. After all, the up part is optional. The down part is not.