I haven't been writing much on handloading recently, but I had to answer a question on my HL list recently, and thought I'd post it here for fun.
I was asked why the maximum listed loads in reloading manuals, supposedly well-researched and tested data, are possibly unsafe in certain applications. Here's the response-
1) Max loads in reloading manuals are based on pressures achieved in/with their equipment, and subsequent to SAAMI max pressure recommendations for a given cartridge.
2) The pressure testing equipment is also subject to SAAMI specs to ensure it will produce consistent pressures within a certain set of variables.
3) SAAMI pressure recommendations can change over time.
4) Pressure-testing techniques and equipment do change, sometimes radically, over time.
5) When making load recommendations, we are dealing with stochastic (not deterministic) models. [Example: 2+2=4 is a deterministic model. yhat = beta-naught + beta-one-x-one + error is a stochastic model. The error term contains a random factor that we can't control, and changes the yhat each time the equation is run.)
6) The stochastic model for pressure for a given cartridge is thus written in terms of a confidence interval, where there is allowed a small amount of error. (This error-allowance is critical...the only way to ensure no error is not to load the ammo at all.) The confidence interval varies based on how much the error term varies in a given situation (i.e. the sample variance), and a constant based on the bell curve determined by how much error we are willing to allow.
7) Rarely, some firearms do not meet SAAMI spec and thus will perform outside the parameters of the testing equipment and produce excessive pressure with a listed safe load. This is why all load manuals repeatedly insist that we start 10% below the listed max and work up.
8) In addition, certain firearms designs are not as robust as they need to be to absorb full-power, maximum-pressure loads over time. Thus, rather than a catastrophic failure, we will find a chronological failure with the firearm. This is often manifest as excessive head space in a rifle, loosening of lock-up in a revolver, or gradual frame damage in a semi-auto pistol, for examples.
9) There are uncontrollable factors based on the skill and quality control of the individual hand loader that can change things in a hurry (i.e. seating bullets too deeply, or not crimping and thus allowing bullets to be pushed into or pulled out of a case during firing, etc.).
So, the likelihood of a catastrophic failure in any firearm when using a maximum listed load is very very small, but still a possibility. The likelihood of a chronological failure when using maximum loads is still small, but larger than a catastrophic failure. So one can generally say that load data in manuals is safe, but one must allow for exceptions and follow the 10% work-up guideline religiously to eliminate (as much as possible) the random factor in the hand loading process.
Bottom line- we rolls our dice and moves our mice, and hope nobody gets hurt.