Hey,
The main problem in the new file I think is that even though your 3 event flows are 'separate', they're still all part of the same over-arching PhysX sim....so they are still interacting and thus affecting the outcome of each other (thus breaking the perfect loop). And by 'interacting' I don't mean colliding with each other....it could be more subtle things, like numerically biasing the results of each other due to the order of collision processing having cross-influence within the global collision solver algorithm...so not anything you can remedy by just tweaking flow settings.
The easiest solution, since they're just 3 offsets of the same setup...is to cache one of them and then just duplicating the resulting tyCache, offseting the start time in each one to match your loop timings. That's how I did many of the loops on the frontpage of the pro.tyflow.com site.
The main problem in the new file I think is that even though your 3 event flows are 'separate', they're still all part of the same over-arching PhysX sim....so they are still interacting and thus affecting the outcome of each other (thus breaking the perfect loop). And by 'interacting' I don't mean colliding with each other....it could be more subtle things, like numerically biasing the results of each other due to the order of collision processing having cross-influence within the global collision solver algorithm...so not anything you can remedy by just tweaking flow settings.
The easiest solution, since they're just 3 offsets of the same setup...is to cache one of them and then just duplicating the resulting tyCache, offseting the start time in each one to match your loop timings. That's how I did many of the loops on the frontpage of the pro.tyflow.com site.