After performing sliding law coefficient inversion, I usually relax the model with transient run and allow it to reach a new steady state before running any forward simulation. I notice that sometimes the relaxed state may be quite different from what was inverted for (e.g. velocity changes over 20% and/or ice thickness changes over 100 m). The inversion itself is quite decent (an example using Helheim see below). 
Is there anyway to mitigate this spurious drift? I tried a) solving stress balance again with the inversion velocity b) set isgroundingline=0 c) forcing basal melting rate. The former two made no difference; the solution c) does work but to find that proper basal melting rate is challenging.
Any tips / tricks on this matter are appreciated!
