I've been working on an amp for a powered speaker using the TPA3255. I have it configured in 2xBTL mode, one output for the woofer and one for the tweeter. We are limiting the outputs using a DSP preamp well below what the amp is capable of. We are almost done with the engineering stage and nearing production, but still with some long term testing to do. Over that time I've probably tested 10 units or more and never had any power amp failures. I started some long term testing yesterday and one of the units *that had been working" failed quickly. The supply still works perfectly and there are no damage to any other part on the pcba. The supply (+50V) pins are burned at the IC and all of the outputs are shorted to ground. I don't see any cause for this kind of failure and that scares me this late in the design cycle.
Any ideas?