Convergence rates are fine (except for sand where the analytic solution is basically a sharp corner). Onset match with the solutions presented in the paper https://doi.org/10.3389/fpls.2020.00316 after fixing the domain size as reported here: https://github.com/RSA-benchmarks/collaborative-comparison/blob/master/C1%20Coupled%20problem%2C%20static%20RSA/C1.1%20Single%20root/C1.1%20Benchmark%20problem.ipynb.
There is a readme describing the cases a bit.