Our CI system runs a pull request poller which checks for new PRs by looking at the PR description to decide which Jenkins job axes to run. We have quite a few axes (QA_JTS_JACORB, QA_JTS_JDKORB, QA_JTS_OPENJDKORB, QA_JTA, XTS, CORE, BLACKTIE, RTS, mysql, postgres, db2, oracle, TOMCAT, JACOCO, LRA) and if the user, say, just wants to run the XTS axis then CI will start a job for each axis. All but the XTS axis will then check whether it needs to run and then terminate. Each axis can end up taking 20 minutes or so. It would be good to minimise this overhead.

For example, all of the QA axes could be run in one job.