test_ff_stokes_channel_3d fails with different velocity

Bug report test_ff_stokes_channel_3d failswith different velocity

What happened / Problem description: !1637 (merged) changed the handling of inflow/outflow BC and adapted the reference solution, yet the test fails on buildbot with

Data differs in parameter: velocity_liq (m/s)_1
 Difference is too large: 199.76% -> between: -1.65925e-12 and 1.6632e-12
 Info for velocity_liq (m/s)_1: max_abs_parameter_value=6.83831e-12 and min_abs_parameter_value=0.0.
 For parameter velocity_liq (m/s)_1 a zero value threshold of 1e-12 was given.
 
 Data differs in parameter: velocity_liq (m/s)_2
 Difference is too large: 127.12% -> between: -4.17537e-13 and 1.53955e-12
 Info for velocity_liq (m/s)_2: max_abs_parameter_value=1.89138e-12 and min_abs_parameter_value=0.0.
 For parameter velocity_liq (m/s)_2 a zero value threshold of 1e-12 was given.
 Test #17: test_ff_stokes_channel_3d ...............................***Failed    1.62 sec

How to reproduce it (as minimally and precisely as possible): Dune and DuMuX master, compiling and running the test.

Anything else we need to know?: Maybe @melaniel can state the version & compiler details she used when generating the reference. I will be trying to reproduce a result tomorrow morning, too. Maybe it is compiler/machine dependent.

Environment:

• Dune version: master
• DuMux version: master
• Others: gcc 5.3 & gcc 7.4.4 (but with vel_1: 196.66% and vel_2: 135.84% difference)