The code for the error propagation in PhotoCalTask and colorterms.py seems fishy to me, and is not really tested by the existing tests. We should work out the math for propagating these errors, implement tests of it, check that the code behaves correctly, and redo it if necessary.

In particular, the interaction between Colorterm.propagateFluxErrors() and calib.absMagErrFromFluxErr() needs to be checked.

We're not helped by the fact that the ref flux errors were originally totally ignored and/or faked (see DM-2308), and reference catalogs weren't guaranteed to even have flux errors (until RFC-535).