Hi Khalil,
First things first: I take it that it is the exact same process, same flow and everything, just triggered differently?
Is this a queued process? Does it run once per day? In that case, when triggered manually, might the load of data be smaller since the big load was already processed earlier in the day?
Is it possible that the SE runs simultaneously with another SE?
I hope one of my questions might steer you in the right direction. If not, could you provide more info?