Bending beam benchmark breaks towards the end when reaching steady-state

[gassmoeller]

As reported by @egpuckett:

When Mack runs this https://github.com/geodynamics/aspect/tree/master/benchmarks/viscoelastic_bending_beam 'out of the box'; i.e., no changes to the parameter files he gets this: https://drive.google.com/file/d/1HaAyWT3zs14i2SnvYIUqAINiaiU_JWOO/view?usp=sharing

The beam quite literally explodes after it returns to its initial state. There is really no other way to describe what happens.

My guess is that we are missing a safeguard against vanishingly small strain-rate (maybe viscosity becomes infinite somewhere). We should check if this is specific to the benchmark case, or also a problem in other models.

[tjhei]

Or, as pointed out by Cedric, dt->infinity as there is no velocity when the beam comes to rest.