Georgi,
Google; graph electromagnetic wave - and you can quickly get pages of graphic illustrations that show the typical planar reversal of the signal (do the same with; electromotive induction - and refresh). But this is precisely where you shine. You have persistently grappled with the quandary of whether what we observe is a product of the characteristics of an observing system or observer. So is the reversal the physical reality or a product of the experimental apparatus and proceedures?
I've been head wrestling a concept of a soloton wavetrain model of EMR for 40 odd years, and there is constantly some observed relationship between interactive fields that presents a damned if it do, damned if it don't scenario. It makes it impossible to argue against the QM methodology because its fascinating just how incredibly complex the behavior is of the humble photon. How does it DO that?!
Experimentally, quantum engineering introduces the trust factor; do we know what we are doing? but we do know that technologically we have highly refined the capability to achieve consistency. In a typical Bell-Aspect experiment, if both polarizers are in parallel, in a run of 10,000 flashes there will be a small but significant incidence when only one or neither detector will register, so we must assume that when the polarizers are at right angles there will be a physical inconsistency of results also. The theoretical ideal is that the Quantum Correlations are the probabilities we could predict IF there were a direct physical connection between the two simultaneous detections. But that's impossible from what we can reasonably explain today. :-0 jrc