create a workflow with the following DSL

stage concurrency: 1, name: 'doit'
node {
 sh 'sleep 600'
}

Start a build, then immediately start another build
delete the first build by going to job/1/doDelete

expected results
the second build that was waiting for the first to finish the stage would progress

actual results
the second build is still waiting for the build #1 to finish - and this will never happen.