Inconsistent quota size computation when documents created then quickly deleted in bulk

How to reproduce:

• Install addon nuxeo-quota
• set log level of org.nuxeo.ecm.quota.size to DEBUG in log4j.xml
• truncate tables document_size_statistics and document_size_statistics if already exist
• Start the Nuxeo Platform
• create workspace CMISTEST workspace
• navigate to ADMIN > Quotas Statistics, tab Compute Initial Statistics and trigger initial computation for both quta updaters
• download to your workstation attached file cmis-test-creates-deletes.sh and testFile.bin
• Execute the attached shell script with the following command (10 threads creating and deleting documents+files in bulk), it creates massively File documents, attaches file testFile.bin (see attachments) to them and then deletes them:

  for i in {1..10}; do ./test-cmis-creates-deletes.sh & done
  

• After 1 minutes of execution, stop script execution with the following command:

  touch test-cmis-creates-deletes.sh.stop
  

Expected result:

totalSize, Live Documents Size and Archived Versions Size should be 0.

Actual result:

totalSize, Live Documents Size and Archived Versions Size are greater than 0.

curl commands to retrieve quota:

Command:

curl -u Administrator:Administrator -H "Content-Type:application/json+nxrequest" -d "{\"params\":{\"documentRef\":\"/default-domain/workspaces/CMISTEST workspace\"}}" http://localhost:8080/nuxeo/api/v1/repo/default/path/@op/Quotas.GetInfo

Result:

{"entity-type":"org.nuxeo.ecm.quota.automation.SimpleQuotaInfo","value":{"totalSize":161480704,"innerSize":0,"maxQuota":-1}}

Command:

curl -u Administrator:Administrator -H "Content-Type:application/json+nxrequest" -d "{\"params\":{\"documentRef\":\"/default-domain/workspaces/CMISTEST workspace\"}}" http://localhost:8080/nuxeo/api/v1/repo/default/path/@op/Quotas.GetStatistics

Result:

[{"label":"Live Documents Size:77 MBytes","data":80740352},{"label":"Trash Size :0 KBytes","data":0},{"label":"Archived Versions Size :77 MBytes","data":80740352}]