Never seen that before.
Silicon has been known to emit light under avalanche (reverse bias) conditions. For example, a 2N3055 emitter-base junction emits a green-yellow sort of light, very faint (it's barely visible indoors at 1A!).
Possible you've got a strange substitute, like a stack of LEDs. Though it should be a lot brighter if that were the case, and it wouldn't conduct in the "forward" direction.
Testing a 1N5242B right now, I don't see anything.
Tim