Some tests are failing in aMC@NLO runs. What is going on?

For some processes (typically processes with jets and/or at high center-of-mass energies and/or when jet cuts are low), it may happen that the aMC@NLO run stop right after compilation, with the error message
"Some tests are failing, run cannot continue"

The concerned tests check whether the soft and collinear behaviour of real-emission matrix-elements matches what comes from the local counterterms. As the limits are taken numerically for 100 randomly generated phase-space points, some instabilities may arise (in particular in the cases mentioned above).

In order to have the code not complaining about them, you can edit the SubProcesses/symmetry_fks_test_ME.f file
and change the maximum allowed fractions of failures (line 122)
      parameter (max_fail=0.3d0)
to some larger value (0.5/0.7 should typically work and be safe).