Stick to the provisors
With the best will in the world, the removal and refitting of a taper chain set is not something that can be repeated infinite times, it just can't, and you know that. Each time introduces a minute deformation of the materials (and let's be honest, the softer aluminium of the chainset is the sacrificial material here, not the hardened steel spindle) until eventually the chainset will 'bottom out' on the taper and not be secured by the retaining nut/bolt. Even under laboratory ideal conditions where the chainset taper interface is scrupulously clean and fitted to the book torque this is not an unlimited process and the described mode of failure will eventually occur. It may be hundreds of cycles down the lifetime of the components, but it will happen, and this time is hastened by frequent unnecessary repetition.
Also, we have to factor in the lack of adherence to the ideal procedure. I work in an industry where machines are serviced by hours and, to pick an example, there are tools that may be serviced at 1000 HR intervals. This involves a fairly intrusive strip down and I can usually tell who did the previous service by how stupidly tight the fasteners are. This is a high-vacuum machine, it pretty much holds itself together once pumped down and the bolts are there just to keep things in place until the vacuum forces take over, but this doesn't stop highly paid and expensively trained 'engineers' from tightening the fittings way beyond any documented values.
Do we think the average bike shop or DIY mechanic can resist exceeding the specified torque for chainset bolts because they 'don't want the cranks to fall off'?
For all intents and purposes, taper chainset removal has limited repeatability. If someone is careful and sticks to best practice this might be hundreds of times. In less ideal conditions, but with some mechanical sympathy it will be double figures. Some ham-fisted feckwits will achieve single figures!