Over in bug 515211, we are crashing inside jemalloc, apparently trying to use an invalid chunk pointer. It would be great to be able to set skidmarks to find out where that chunk comes from. (There are two different functions that might be returning the chunk, and other things about their operation might also be of interest.)