Acceleration and Velocity are vectors. Changes in a velocity vector are an acceleration. Therefore when photons change direction technically it’s a form of acceleration.
I thought photons are always moving in straight lines from their perspective, and it's space that's bent. Unless it's through a medium, then they just get absorbed and re-emitted, sort of.
Ok but photons don’t change direction either. Treating photon scattering as an individual particle accelerating due to an applied force, well that’s just not a correct description of how perturbative QED models photon interactions.
Since photons are indistinguishable, it's hard to say too much concretely, but it some sense a diffracted photon is different photon. In order for a photon to interact with say, a diffraction grating, the interaction is done with "virtual photons".
So for a photon to change course, aka accelerate, it does it by absorbing a virtual photon and emitting another. Whether that is the "same photon" after the interaction is kinda more philosophy than physics, at least to me.
Feynman diagrams are surprisingly accessible for how much information they contain. It's one way to think about photon (and other particle) reactions.