I'm working on a flowline marine ice sheet model looking at some retreat scenarios. I'm using the mismip bedrock geometry and have the ice sheet initialised so that it is stable just on the seaward side of the bedrock bump. I'm using Coulomb friction at the base, the lateral friction function, increasing the glacier width to decrease the buttressing and a separate solver for melting below the ice shelf (SubShelfMelt solver, its in the github repo but not documented on the wiki).
The problem is when the grounding line retreats and seems to get stuck just at the point where the bedrock should start to be backwards sloping. The output starts to look quite strange with many "Retreat of the Grounding LIne" type messages within a timestep, but then the GroundedSolver at the end of the timestep puts the grounding line in the same spot it was at the previous timestep. I also see a lot of "!!! Ne <=0, nodenumber 2 -1.728410418593285E-003" type messages, suggesting something strange going on at with the effective pressure around the grounding line and on the back slope.
Also, if I restart the simulation from one of these timesteps then I sometimes get a huge jump in the grounding line retreat on the very first timestep.
Has anyone come across this behaviour or have any suggestions about what I'm doing wrong here? I'll attach my sif and a snippet of the output. Any kind of help would be appreciated.