In physics we use a convention that light goes the same speed in every direction. It was verified experimentally that there is no preferred frame of reference and no anisotropy of speed of light was detected.
However these experiments always measure two way speed of light, as it is said there is no way to measure single way speed of light. Ie light might be going one way faster and slower the other way back, averaging to c. It all makes perfect sense and all if we consider closed system experiments, but if the light didn't had different speed depending on direction on the same axis, shouldn't that be clearly reflected on CMB? In most extreme example where light goes 2c one way and instantaneous the other way, we shouldn't see any cosmic background radiation from that side as it already passed us long time ago. What more, there shouldn't be any observable horizon and we should see as many stars as there is in the total universe and no matter how deep we look everything should look no younger than our part of universe. In general, any difference in one-way speed of light should be reflected on the cosmological horizon.