On some jobs, I use a condition set to check the output of the previous task, and send an email notification based on the result. Randomly, although the condition set is set up properly, the email was sent, even though based on the conditions, it should not have sent.
I've noticed that when I watch this type of job run, sometimes it looks like it takes a fraction of a second longer for the output to populate in the window of the GUI than it takes the task to finish. My theory is that perhaps the condition set in the next task is being checked just before the output of the previous task that is it checking is fully available. I found a way to force the job to pause before the output is checked by the condition in question; but since this happens randomly it's difficult to test, since I don't know if a job is working correctly because of the pause, or because that situation just isn't occurring at that job run.
The other interesting thing is that the email task with the condition set uses the same output that the condition set is checking for in the text of the email. Even though the condition set falsely reports that something is missing, the output in the email shows everything was present that the condition set was checking for.
I'm interested in knowing whether you think that is plausible, that the condition set may need a bit more time before reading the output of the previous task, before I set up all of our jobs based on that assumption. (This is most frequently used with jobs that download files from an external - customer - FTP site - it's the output from the FTP task that the condition set in the following task is reading - so perhaps there is some latency involved.)