100nF MLCC bypass capacitor spontaneously exploding on 3.3V bus, how does this even happen?

Could be high ripple currents (or poor cooling and moderate ripple currents) causing overheating. Or it could be one or several of the mechanical stress/defect sort mentioned in the comments, or a combination of such factors. Without knowing if the capacitor is prone to blowing up on more than one copy of the board it's hard to say for sure. Post-mortem analysis of a single failure without extensive supporting data is not definitive.

Linked below is a TDK app note hosted at Digi-Key which pretty much explains that while MLCCs are often not exactly ripple-current rated, ripple currents do affect them - but it's more of a temperature thing than a specific current. They also mention that when ripple currents are measured on MLCCs it's typically at room temperature - so it's quite possible that some lazy designer found a number, ignored the "25°C" associated, and said whoo-hoo, here's my ripple current - let her rip - at 65°C.

http://www.digikey.com/en/pdf/t/tdk/ripple-current-mlccs

The "4V capacitor on 3.3V supply" (also mentioned in comments after I wrote the above) is likewise a poor design choice that may well contribute, but it's unclear if the spec is what the actual capacitor is, since you indicate a source that is evidently not the manufacturer of the board (chipmaker? in which case they are nuts to spec that) for the spec.