I'm wondering if someone has ever done an experiment with the ability to check whether a large cohort of photons is seen (or not seen) to gravitationally affect other photons, which would be definite proof (one way or another, once the accuracy is high enough) of whether the photons traditionally massed-as opposed to merely space-time bent.
That there is always a question of accuracy makes it impossible to have a "definite proof". All we can do, is establish an upper limit of photon mass, which can be done in experiments like this:
http://www.princeton.edu/~romalis/PHYS312/Coulomb%20Ref/TuCoulomb.pdf . Check pages 4 and 8 for lists of all the experiments ever used. You need to have a good grasp of higher maths as well as physics to be able to grok how it's done though.
Ultimately, you're always going to be limited by the uncertainty principle, so you'll never really know for sure, no matter the accuracy of your measurements.
http://www.scribd.com/doc/62099162/22/Dispersion-of-starlight describes experiments(starting with De Broigle's in 1940) like the one you have specifically proposed.
edit:
I like your examples though. They highlight what I was trying to say. Everyone knew that light functioned like a wave, but suddenly it didn't. People probably went "Why is everyone always trying to prove science wrong?" regarding that too, when the truth is that questioning if it really does work like that is what makes science science.
You know we(me, you, Virex) are all on the same side of the barricade, right?
Nobody is advocating not questioning science. What is being advocated is moderation in how it's done. In the light-as-a-wave example, you could have three basic reactions:
1.Science is religion, ergo light cannot be made of particles.
2.Science is wrong, ergo light cannot be a wave anymore.
3.Science changes to fit the data, ergo we've got wave-particle duality to describe behaviour of light.
What is being said, is that 2 is just as radical an approach, and just as bad as 1.