The records will be rotated once replaced by new provisioning attempts. There is a way to improve this as you suggested by detecting cloud deletion but we can not detect cloud rename (and tell renames from deletions/creations) so I prefer to keep this the way it is. Another thing is we can not detect this for templates that are not covered by Cloud API at all.

I am keeping this open as I am not quite satisfied with the way things are. Consider last cloud is deleted and the records keep hanging there forever.

Thanks Olivier - I will keep watching this issue.

Hi Oliver,

We have provisioned hundreds of new docker containers but the old Dockerhost still remains. Any specific way to manually remove it? 

I know something like that can happen when the activity is not considered completed - presumably because it has leaked in PROVISIONING state. The easiest way to go from here is to edit the .xml file while Jenkins is down and remove the offending entry. The fix for this bug is on my todo and will hopefully be released soon.

We are seeing this issue as well. Is there still plans to resolve this?

My workaround:

I tried to delete some of the xml elements which should be removed in in org.jenkinsci.plugins.cloudstats.CloudStatistics.xml but the Jenkins crashed after login. 

So I came up an idea to delete everything in that file and restart Jenkins. It didn't crash. And all the cloud statistics history is gone. I'm OK with this because what's matters most is to display the latest provisioning status and remove the annoying old cloud status which should have been removed long ago.