Reason

From Joe Kelley at the NRC:

I am working on a 7-cell model that uses the GapHeatTransfer BC for every pair of sidesets as was done for the "reference" calculation in the extended scope task.

In that calculation, there were only two axial regions in the meshing-tools script, the support plate and everything else.

The Poisson solve absolute tolerance sets the accuracy of the Poisson solve. Given the value of abs_tol, the error level targeted in the Poisson solve is abs_tol * 4 * pi * G * rho_max, where rho_max is the maximum density on the domain. For Cartesian simulations the default value is 1.e-11, and for non-Cartesian the default value is 1.e-10.

At high resolution, and/or for complicated mass distr

If point-wise source terms are sufficiently performant, we should only use those since they separate DG internals from equation-specific stuff. However, for this decision to make properly, we should run some 2D and, more importantly, 3D tests.