Solver Penalty resets to 0.0 in MPI Parallel runs involving Velocity BCs (Moving Wall) #746
Description
Hi there,
I am encountering a critical issue when running UWGeodynamics models on an HPC cluster using MPI (Parallel mode).
When I apply a specific velocity boundary condition (e.g., a "Moving Wall" / Push-from-rear setup) in a parallel environment, the solver's Penalty parameter appears to reset to 0.000000 during the run, regardless of what I set in the Python script using Model.solver.set_penalty().
This causes the solver to fall back to a default BSSCR configuration with zero penalty, resulting in a singular matrix (or extremely poor conditioning). The pressure solve iterations skyrocket (e.g., >390 iterations), causing the model to stall or run extremely slowly.
