I prepended the $PATH with the path as it was from root. However i do not think it has something to do with the problem.
I tried to invoke the “old” workflow (the one i had not edited), and it was enqueued successfully (from non-root and with old tool xmls). Unfortunately i had not tried this workflow between the error manifestated and path editing (cos it requires manual entering of millions of parameters).
Then i tried to invoke some “new” workflows. Sometimes i got “Invocation scheduling failed”, sometimes “Workflow submission failed” Workflow submission failed , and one time i even got successful invocation (but this time i forgot to push “Send results to a new history” button).
I tried to edit the old workflow again. This time i was unable to connect the genome input to MergeBamAlignment even if i created the new instance from the left panel. Workflow automation? - #4 by wormball However after some random mouse clicks the editor allowed me to do this, but only in two of three MergeBamAlignment steps i have. So i had to copy one of these and replace the third such step. After i finished the editing and saved the workflow, it enqueued just fine.
Maybe this step is root of all evil. On the other hand, when i tried to crop the workflow to find the bad step, it started to enqueue perfectly while all three MergeBamAlignment steps were present. Also i suspected that SnpEff eff may be guilty, but again, sometimes it worked with it and not worked without it. So the cause of this error is still a mystery. And the fact that the galaxy says nothing about this (and not only this) error whilst it writes millions of letters to its log makes me a sad panda.