I was trying to see what IBM would do under the hood with a CCNOT gate. Something appears to be erroneous with the run.
The circuit above should produce state 11100 with 100% probability, as confirmed by the visualizer (and the basic definition of the CCNOT). However both the simulator and the actual backend (ibmqx4) give all 1024 shots as 00000. Something is definitely not working right here.