...that the definition of the speed of light includes the term "in vacuo" for a reason?
It means, literally, "in vacuum". As in, the complete absence of matter along the path. Which is impossible even in laboratory conditions, never mind out in space where deep interstellar density runs on the order of tens of atoms per cubic metre. That might not sound like much but with the quantum probability of a photon polarising and forming a pair with mass, that pair will be subject to gravitational effects (even from a dust particle). The vector is changed for the pair and therefore the photon. There is an increasing probability that on its way through a region with stellar-plus-sized masses, the photon/pair will be subjected to immense gravitational effects resulting in a lensing effect to the observer. This does NOT mean that the speed of the photon is changed - it is still travelling in a pretty constantly rare soup - only that its path is changed. Its lens-pair photons will arrive at the same time to the observer *providing they do not encounter another lens on the way* - which given an infinite universe is a definite possibility, however unlikely/unobserved within our observation sphere. Should this happen, I predict a delay between one photon path intersecting the observer and another from the same source at the same instant having passed through a lens. Could this be predicted using a single light source and two sensors: one through a glass fibre path straight to the source and another through a glass fibre path of the same material characteristics but that takes a longer and more circuitous route?