Also, some more food for thought on this statement. I don’t think this should be like that (could be in how UE is solving this, but it shouldn’t).
If we eliminate lag, different tick rates, etc. (which in this case, running on a local computer editor, we are) and you execute the same simulation scaled by delta time, then you should get the exact same results. Again, we are not introducing anything external like the player crashing into another one, or a wall, etc.
What you are saying would be true if you JUST sync the position, and the client needs to slowly interpolate to the next server update position. In order to solve that, you instead sync position AND velocity, and allow the client to run the same simulation, essentially predicting what the next server update will be.
The corrections would only trigger when something happens in the server simulation that the client couldn’t predict, like input from another player, a collision, an impulse changing the velocity, etc.
But for this simple case, I would expect the result to be closer to the second part of the video (coefficient set to 0), but with both positions matching (unless the position not matching has something to do with the frame the debug lines are rendered).