The specifications that are outlined for a product are the limits of that product. Any output of that product that is outside of those bounds is not supported and is not usually a liability to the company. To avoid customer frustration and risk of failures closer towards the limits of a product, buffers are often put in place which, although seems to increase risk, helps in the long run. Furthermore, with my experience in validation, the limits are often the furthest we will go for testing. In addition, however, we dedicate a good amount of time in attempting to bypass those limits which we always enforce hardblocks on (this is for software). Ultimately, testing the supported limits, as well as the blocks enforced to prevent the user from exceeding those limits, is how validation is typically handled, from my experience at least. As far as one vs two sided limits, it of course depends on the situation. Often times, two sided limits are enforced because it is unreasonable and usually impossible to support a product that can have an infinite spec. However, in some cases, infinitely “better” is better. For instance, certain products must have a lifespan of AT LEAST “X” number of years. Obviously, the longer the better, as long as the product can function properly within the defined specifications during that period.