Computational science projects performed at the University of Chicago Flash Center frequently give rise to interesting problems of applied mathematics and mathematical physics. I will describe how three of these came to our attention, and were subsequently solved: (1) The taming of an inaccurate (but widely-used) algorithm for multipole self-gravity in hydrodynamics; (2) The proper description of turbulent convective spectra in core-collapse supernova progenitors; (3) Catastrophic trouble attends implementations of the Biermann Battery effect near shocks in numerical MHD -- I will describe how we found the trouble, and shot it.