We shouldn't test for zeroness using qFuzzyCompare, instead we should
use some predefined epsilon value. This ensures we never try to divide
by an extremely small value, not representable by double precision