asayers
This is NOT a self-proof - it is a meta-proof. Taking arithmetic as an example, a self-proof of consistency would be a derivation of the consistency sentence (ie. "I am consistent", or "¬◻(0=1)") from the arithmetic axioms. That is not what we have here.
The existence of a proof of the consistency of a theory does not put it at risk from Godel's 2nd - for that we require that a theory prove its own consistency.
As far as I can tell, Hewitt begins with a proof that consistency and the validity of proof by contradiction are equivalent, and then proceeds on the grounds that consistency is proven internally to the theory - which it is not.
Frames of reference are fairly obvious in SR. In General Relativity they're more non-trivial.
Say the pen is spun in the box by a classical random number generator. You don't know which way the pen is spinning, but you do know that it's either been spun one way or the other. You might say that the probability of finding it spinning clockwise when you open the box is 1/2.
Now say the pen is prepared in a superposition of the two spin states, again in the box. As before, we might say that the probability of discovering the pen spinning clockwise is 1/2. However, this time we the probability isn't generated by our lack of knowledge: we know exactly what state then pen is in. When we open the box, however, the pen will change instantly to the state of spinning clockwise, or the state of spinning anti-clockwise.
In the quantum case, the probability is an expression of what we think will happen to the pen, not what we think has happened to it.
It is hard to grasp, and harder still to believe. There is, however, good reason for thinking that, sometimes, the pen changed just as we opened the box.
The scenario you outlined is quite important and actually has a name: it's the "Bertlmann's Socks" thought experiment, and if you want to read about it I'd seriously recommend the paper by the great J S Bell[1].
To use the language of your example, both pens begin in a superposition of clockwise and anti-clockwise. It's not the case that each pen has a particular spin value, and that we're simply unaware of which has which. The pens really are in a superposition.
Until, that is, a measurement is made on one of them. At this point the joint pen-pen system collapses and both pens have determinate spin values. The nature of the entanglement ensures that those spin values are different.
The notion that quantum-entangled particles could have well-defined properties that we're just ignorant of was actually pretty popular in the early days of quantum mechanics. In fact, the theory was put forward by Einstein, among others. As it turns out, however, that we can test for this. The tests have been done, and it seems Einstein was wrong on this one.
[1]: J. S. Bell (1980), "Bertlmann's Socks and the Nature of Reality"