If a Cloud Integration Poll or Rake task is "paused" for some time (status changed to Completed), so that its executeAt date ends up >1h in the past, and then "restarted" (status changed to Queued), and then execution starts again on the task (status changed to Executing) the "reset abandoned task" will see that task as "abandoned" (because it looks for tasks in Executing status with an executeAt date >1h in the past) and then change the status to Queued.

Meanwhile, the Poll/Rake task scheduler now sees the task (that is actually still running) as Queued, and might resubmit that task for execution, causing copies of the same task to be execute concurrently. A given task is not meant to have concurrent copies of iteslf executing, which breaks the task handling as one of the tasks will successfully complete and change its status from Executing to Queued but the other will fail and an error event is generated when really the task completed successfully.

The way these "paused" tasks are handled when "restarted" needs to be adjusted so the "reset abandoned task" does not modify the restarted+executing tasks.