Can the spin of a photon change during its "life"? Or is the spin set in one of two possible states at its moment of creation and does not change for the rest of the duration of its "life"?
 A: The life of a specific photon is very short, considering its speed. It mostly ends up on some matter and it interacts with it .
A photon has spin one, always. That cannot change.
It can have a projection of spin +1 or -1 depending on the polarization of the other particles in the interaction that produced it. It retains that polarization until it interacts  with a particle/atom which absorbs it or scatters it. It can change spin projection only through the interaction.
A photon that escapes to the vacuum of outer space without interacting keeps its spin projection until it interacts.
A: The photon's life proper lifetime (that is in it's own frame of reference) is zero so nothing can happen to in{*} outside of the reactions the create and destroy it.
The things that can not happen to it include changing spin.

{*} This is why neutrino mixing shows that neutrinos are not massless.
A: Spin in all forms is conserved absolutely, so the simple answer to your question is no: it cannot drift or change without some external event affecting it.
The tricky part is that due to entanglement, that "external event" could be very distant in space or even time, making the actual spin pretty mysterious.
The (long) addendum below addresses the conservation and entanglement issues in more detail for anyone interested.

Your question entangles two different questions, so to speak
The first question is what happens to the specific polarization (spin orientation) of the photon when it is detected or measured, which because of entanglement can give unexpected answers that may seem to contradict how you know the photon was generated. If you examine only that aspect of photons, it can indeed look as though the spin orientation of the photon has drifted, because any one measurement of a photon orientation may seem quite different from, say, some precise source of polarized light from which it was generated.
The second question cuts deeper, however. That question is this: Is angular momentum absolutely conserved no matter what direction of spin the photon has when it is detected? The answer to that is "yes, and with absolute precision."
What that second part means is that if you detect a photon with one type of spin at a remote site, then if another photon was co-generated to cancel its initial spin, then that second photon will become a photon that on average will still cancel out the spin of the first photon. That's the entanglement effect you hear so much about, the "spooky action at a distance" that if arranged carefully can produce very interesting forms of encryption. (Alas, though, it cannot transmit real information. But that's a separate issue covered by other answers to other questions here.) Entanglement in all forms is always driven by the need to preserve one of the fundamental conservation laws of physics, in fact, so that they remain true even in situations where particles or objects are widely separated.
So, the bottom line is that if you are talking about sets of spins, such as pairs of canceling spins, the answer to your question is "no" for the overall set: You cannot change the net spin of the set, if no external spins are allowed in to entangle things.
But if you are asking only about one photon, it still cannot "drift" or change by itself over time; the wave function preserves its spin state precisely. What complicates that case, however, is that even though spin cannot drift, it can be reset by events elsewhere in the universe -- and possibly very, very remotely, such as at the other end of it, literally.

Finally, I must add this addendum: If you read what I just said very carefully, you will notice that I did not say that the second photon is converted to a spin that "exactly" cancels out the detected spin of the first photon. That's because it too will have some uncertainty when it is detected. This means that if you have a second detector (polarizer) that is exactly aligned with the one that detected the first photon, you can get precise cancellation of the spins, and the universe is happy.
But what actually happen in most cases is more like a reverberation of a bell, albeit one that is very faint indeed because it involves atoms "ringing" much larger physical objects. What happens is this: If you measure the second photon off axis so that it is forced into a result that does not quite cancel the angular momentum of the first photon, you create a second entanglement, this time between two non-quantum objects, specifically the first and second photon detectors. This is pretty much unavoidable, since it's the only way to keep the exact conservation of angular momentum going. And in principle, this process could be continued through even more cycles if you could very (very, very, very) precisely measure the angular momentum of both detectors.
So why don't you hear about this idea of reverberating entanglement? Mostly because it would be so hard to detect that physicists usually just sort of fold the cards as soon as entanglement reached the level of a classical object. After all, a classical object has so many things going on that any such slight excess would be dissipated very quickly anyway.
But not entirely. And therein lies an experiment I've never heard tried: Multilevel entanglement using very small, very cold, still quantum devices as the detectors, so that at least one additional level of unresolved entanglement after the first pair of detections can still be observed.
(If anyone knows of such experiments, please comment. I have not done a literature search on it, but the arguments for it are sufficiently straightforward that I would guess someone has at least proposed it in some paper somewhere.)
