In the attached project there is a process designed with a Multiple Instance Subprocess with one Human Task and a boundary timer event, time set to 'duration' value '10s'.
It is expected that this timer fires only once for every Instance of the Subprocess and that this timer is not active after aborting the whole process. But deploying the process with runtime strategy 'PER PROCESS INSTANCE' leads to the following behaviour:
For every subprocess instance the timer is repeatly triggering every 10 seconds and after aborting the whole process the runtime engine tries to trigger it once more leading to an error in log, the process seems to be terminated correctly anyway.

If you change the dependency to runtime version 6.4.0.Final it seems to work, but not with runtime version 6.5.0.Final.