Dear Waferthin,
Frankly, it all depends on what you mean by "large" samples and "frequent" testing. And possibly on the number of products involved.
HACCP agrees with you that routine micro.analysis of the end product should be minimised. This essentially comes within the verification program.
IMEX, the worst case scenario is that some customers simply demand detailed micro.data on every lot/sublot. Most companies tend to cave in to this request for business reasons.
If you do have freedom of choice, it becomes a question of where to set the verification "balance" between adequate "self-protection" and a waste of time and money. Unfortunately the temptation is to bias the sampling to the former for obvious reasons.
The statistical aspect IMO can be relatively undemanding if a minimalist approach is followed, eg as per EC micro. acceptance standards.
Rgds / Charles.C