I have to say I'm disappointed in the level of discussion on this subject (elsewhere online and here on HN.) It is obvious to anyone who has ever paid attention to an LED that this is a failure of the phosphor adhesion; more specifically, the phosphor (a fine powder which converts the nearly monochromatic blue-purple LED light to the broader white) is held in silicone, in a small dab on the surface of the LED. The fact that this silicone lasted only a few years is the surprising part; sure, the environment inside the streetlights is harsh (large temperature and humidity swings) but a high-quality silicone should be able to handle this for significantly longer.
It would be nice to read some discussion of the core issue - the silicone - to me, it seems like a total failure of the silicone and not just an adhesion issue, as the photos seem to demonstrate small cracks leading to larger cracks and then total failure. Perhaps related to the addition of something new to the phosphor mix, which might change its properties under thermal cycling? Poorly mixed silicone? Unsteady temps as the silicone was applied?
It would be nice to read some discussion of the core issue - the silicone - to me, it seems like a total failure of the silicone and not just an adhesion issue, as the photos seem to demonstrate small cracks leading to larger cracks and then total failure. Perhaps related to the addition of something new to the phosphor mix, which might change its properties under thermal cycling? Poorly mixed silicone? Unsteady temps as the silicone was applied?