For posterity sake to remember the purpose of this: with isothermal EB, or potentially other EB cases, the chi_correction in the SDC iterations was not converging. This was because the redistribution process is arbitrary and does not enforce that RhoRT is maintained in the redistribution. We do a lot of work to ensure that the MAC velocity field divergence will result in constant RhoRT, then throw that away by applying redistribution to scalar advection. @dmontgomeryNREL found that turning off advection redistribution resolved our problems and did not lead to instability.

Case setup:
Rectangular domain with isothermal cylinder EB initialized with pure N2 at 143 K. The inlet CO2 mass fraction and temperature slowly increase from $Y_{CO2} = 0 \rightarrow 1$ and $T = 143 \rightarrow 293$ K, while the inlet N2 mass fraction decreases from $Y_{N2} = 1 \rightarrow 0$.

These plots demonstrate the effect of turning off flux redistribution for advection. The first plot is from a divergent simulation that resulted in NaN's with advective fluxes treated with state redistribution. The second is from a simulation where the redistribution is turned off for the advective fluxes.