Regarding the HQL query that is used to calculate nextDueDate in DispatcherThread, JobSession.getFirstUnownedDueJob, I just checked with a Hibernate expert and it could indeed be the case that this query returns the bigger part of the JBPM_JOB table.

The HQL query is defined like this:

  <query name="JobSession.getFirstUnownedDueJob">
    <![CDATA[
      select job
      from org.jbpm.job.Job job
      where job.lockOwner is null
      and job.retries > 0
      and job.isSuspended = false
      order by job.dueDate asc
    ]]>
  </query>

and the result is handled that way:

return (Job) query.setMaxResults(1).uniqueResult();

Hibernate can issue a request that returns all records from the DB, and then just restrict the result to a single row. To prevent this potentially expensive DB call, the 'correct' way would be to add further restrictions to the HQL.