Brian's quite right in the description of how it's derived, but the regs don't give the underlying rationale. I've promised my doctor that I wouldn't get sucked into the discussion on defining limitations but, like fine wine, I just can't resist it, so...sorry.
Designer chappies are tasked with meeting a design specification, that usually includes specific requirements like max speed, g/structural limits, etc. In the interests of self-preservation - particularly in today's litigious society - and under the usual commercial pressures that want to give as little away as possible, he will design to those limits and no more. When built, the prototype will be tested to those limits and no more - done the job, now pay the bill. Depending on the company, the project itself, the certifying regulator and/or the customer, it may or may not be tested further to demonstrate an acceptable margin above that limit that caters for in-service variations and different pilots' ability to respect the limitations. Whatever, the company and/or regulator doesn't want anyone in service going above that limit - proven with a margin or not - so the resulting limitations for service use come out with a plethora of additional limits designed to satisfy the commercial requirement for demonstrating spec compliance while keeping the day-to-day service user away from them. (In civilian aeroplanes this is not usually that much of an intrusion, but it sure as hell is in military aircraft where the difference betwen "Normal Operating" and "Never Exceed" limits often represents a significant reduction in operational capability.) In civilian terms, this difference is represented by Vmo and Vne.
Call me old-fashioned, but I was taught - and have applied as rigorously as I could during my career - that flight testing was there to demonstrate that the full specified flight envelope could be achieved by the average operating pilot within an acceptable margin. So, for example, if the max specified speed was X00 kts then I would expect to demonstrate it to X00 x Ykts, where Y is a factor that is dependant on the pilot's ability to observe and respect the limits. (This factor varies with aircraft characteristics and the limit in question - for example, if the aircraft has no natural stall warning then it is necessary to allow a large margin on any defined pilot-monitored angle of attack limit, because the aircraft provides little cue to help him.) In my experience, this philosophy is not followed by the engineering and commercial specialists - I've tested aeroplanes to their max speed limit and demonstrated that the wings don't fall off, but I've been prevented from testing them further: instead, the in-service limit has been reduced by 10% or so to provide a crude additional margin.
The result: in-service aircraft that are not cleared to their full potential, which (in the military case) crews then take to war and find out empirically what they actually can do - what an amateurish way to work!
OK, I've managed to restrain myself enough to remain polite on the subject, best I now go and lie down in a darkened room and sip camomile tea.