It appears that the kubernetes pods that are spun up to run the pipelinse when using drone-runner-kube are not canceled if a user cancels the build from the UI. Although they appear canceled from the UI they continue to run until the have completed.
This is actually a relevant issue, as the processing seems to go on an on. For very resource-intense builds, this actually takes away available resources from other builds.
Both Brad and I tested and were unable to reproduce this issue. We encourage anyone that is able to reproduce to triage the code and send a patch to: drone-runners/drone-runner-kube