A test shell script is invoked once for every task setting the TASK environment variable, so running a fixed set of tasks from it breaks running tests in parallel when this happens twice or more at the same time.
Instead of the shell script I've added a different workaround using an extra sby file an inline python to run the non-default tasks and to handle the expected failure of the nofullskip task without changing the example sby file.
A test shell script is invoked once for every task setting the TASK environment variable, so running a fixed set of tasks from it breaks running tests in parallel when this happens twice or more at the same time.
Instead of the shell script I've added a different workaround using an extra sby file an inline python to run the non-default tasks and to handle the expected failure of the nofullskip task without changing the example sby file.
cc @KrystalDelusion