Description of problem:
Running a continuous smart state analysis workload that scans several VMs, a single Host, and a single datastore in VMware, the total number of threads for each smart proxy worker appears to grow until the worker is recycled.

Workload: Via UI, every 45 seconds, scan 3 vms, scan a host, scan a datastore for 2 hours.

Version-Release number of selected component (if applicable):
5.5.3.2
I am unsure if this affects previous versions but will run a test on an older version to see if it exists.

How reproducible:
It reproduces on a VMware environment.  Running against my RHEVM environment it does not reproduce, however the RHEVM workload does not scan data-stores.

Steps to Reproduce:
1. Add VMware provider to appliance
2. Add vddk to appliance
3. Continuously scan vms, host, and datastore and watch number of threads for each smart proxy worker.

Actual results:


Expected results:


Additional info:
Perhaps it is related to https://bugzilla.redhat.com/show_bug.cgi?id=1321934

Attached are screenshots of system performance during 2 hour workload.