so it looks like I may have found something...
One theory I have, has to do with the logic for detecting if a process is still running...the plugin executes a "ps" inside the container while the main script is running...the result of the ps call is harvested from the output of the call and used to verify that the main script is alive...BUT it also exports all the env vars BEFORE running the "ps" command ... and some env vars may not be correctly escaped therefore causing some stream corruption between output and error and the EXITCODE is not harvested and thus it thinks the process is not running hence the "-1"
I found these traces in my logs:
WARNING: Unable to find "EXITCODE" in a valid identifier
WARNING: Unable to find "EXITCODE" in r kill -l [sigspec]
I took the combination of https://github.com/jenkinsci/kubernetes-plugin/pull/232 and https://github.com/jenkinsci/kubernetes-plugin/pull/218 to produce a new SNAPSHOT.
Have not had a "-1" since.