Changing Cloud configuration while there are existing containers seems to really screw things up.  I changed the docker host URI (to a different host) and changed the Container Cap.

As a result, on the new host, I had a disagreement between the Container Cap and the actual container count.  For example, I had the following combination at one point:

• Containers running: 1
• Container Cap: 7 (I kept increasing it, but it only helped temporarily)
• Log shows: 
  Sep 17, 2018 7:16:35 AM INFO com.nirima.jenkins.plugins.docker.DockerCloud provision
  Not provisioning additional slaves for container&&centos&&mem32; we have 3 executors being started already

This did not change in 3 days (Jenkins just kept printing this in the log), so I hacked my way into clearing the count of containers in progress in a jenkins script console: (I have only one Docker cloud)

Jenkins.getInstance().clouds.get(com.nirima.jenkins.plugins.docker.DockerCloud).CONTAINERS_IN_PROGRESS.clear()

And now I'm getting the right count of provisioned containers!

Incidentally, the content of "CONTAINERS_IN_PROGRESS" was a map with two keys (old and new Docker host), new one was set to 3 and the old one was at 7...