Yeah, we thought about that, but I'm not sure how stashing would work for parallel running stages, also we would want to minimize network traffic between the nodes.
Right now I'm investigating where the artifacts are stored and hacking something in shell that could merge the results from different stages.
(AFAIK the artifacts themselves are stored on the master in the $JOBNAME/builds/$BUILD_NUMBER/archive.zip, so I'm playing with that file)
Also we are using a slightly outdated version (2.121.2), update is in the works, so that may solve the problem?
BR,
Laci