Hi Klemen - it’s certainly unusual and most of the time we’d expect only one submission. The reason we don’t prevent it is that it’s possible (though again unlikely) for multiple “isntances” of a flow to need to run simultaneously — for example, if multiple tasks all request retries at slightly different future times. Coordinating that activity is a challenging distributed systems problem 😅 and sometimes results in behaviors like the one you saw. However, I’m flagging the ID for our team and we’ll see if we can identify the circumstances and if there are additional steps we can take to limit this repetition.
One thing I already know is the upcoming release of Core will improve what we call “version locking” which is our way of coordinating activity among potentially multiple distrubuted agents. Those improvements plus some changes we’ll enable in Cloud will reduce the possibility of this further.