Saturation values above 1
GEOS formulation allows saturation values to go above 1 (not sure about below zero). The computed saturation values are used as is to compute rel perms that might be not really physical thing to do and also may mislead the solver blocking some flow or allowing too much flow. More reasonable way would be to normalize saturation values used for rel perm computes, see, e.g., https://drive.google.com/open?id=10dHD58pO8dy5hSz4RjVo8m0Mjs6C0a69
Attaching small test case. The first nonlinear steps for this case show saturation changes of more than 1 indicating that (at least the intermediate) saturations are greater than 1. caseFiles.tar.gz
Should be addressed by new formulation in https://github.com/GEOS-DEV/GEOS/pull/3443