@jedbrown Indeed, upgrading the #RiemannSolver to #HLLC (which includes a contact wave) from #HLL (which only estimates the acoustic waves) fixes this problem. Here's a pair of oblique bubble simulations to campare HLLC (left) with HLL (right). Unlike most cases with finite volume methods (which use Riemann solvers at every grid interface), a high quality solver is important for continuous Galerkin FE boundary conditions.