The way to think about the wave behavior is that a wave is a large collection of photons, each of which travels through an optical system as a wave but interacts with a detector as a particle. Let’s suppose you set up three detectors to measure whether a photon is transmitted, reflected, or turned into a plasmon. If you send photons through the system one at a time, you will only detect each photon at one place. When you combine the results from a large number of photons, however, the results build up a wave pattern. I probably explain this more clearly in my post on Young’s experiment.

From a particle view you will observe one chance in three for every measurement, using three ‘observers’?

But for a wave all three happens when measured?

(This is a little oversimplified, because the wave produces interference effects, but roughly captures the idea.)

As if each photon ‘wavepacket’ interacting, in its ‘mediation’ get split up into three components, superimposed too? And if they do, is that the way I can expect all light to act when ‘interacting’?

As I said, it’s a weird question.

For the case of a hypothetical “negative index material” (NIM), not only does conservation of momentum seems to be problematic (as shown in this post), but the closely-related phenomenon of “time reversal” [J. Pendry, Science 322 (2008)] is no more intuitive. The demonstration using Snell’s law implies that transmitted waves in medium 2 are homogeneous (i.e. not evanescent like plasmons). Thus we can show that Snell’s formalism at the vaccum-NIM interface requires enforcing “negative time” in order to yield a transmitted plane wave with a proper wavefront.

Furthermore, a recent theoretical investigation points that a NIM would violate the second law of thermodynamics if it existed [V. Markel Optics Express 16 (2008)].

Hence the existence of a “true” NIM still raises a number of fundamental questions IMHO; meanwhile research for negative refraction and related devices is still going strong.