Strange residuals behaviour
 

Hello everybody.

I'm trying to run a very simple case: a closed volume of water (cylindrical geometry) in which a disk rotates at 303 rad/s around z-axis.
I'm using OF211 and MRFSimpleFoam as solver (MRF steady state solver).
The geometry is very simple and I built a structured mesh (hexa block mesh) of a quarter of the cylinder.

The BCs are:

- rotating disk (including shaft): no-slip walls
- casing : no-slip fixed wall
- cyclic AMI ( = ggi) faces

The case is steady state, k-epsilon turbulence model, I order numerical schemes.
For foam users, I attached the case.

I first run the simulation for 3500 iterations and all seemed "normal ".
Then I restared from time step 3500 to 8500 without changing any thing and the residuals shown the strange to me behaviour that you can see in the picture I attached.
First they seemed to converge, than they grow again and keep higher than before.
The point is that, although this strange behaviour, the fluid fields (velocity, pressure) did not show any important change between before and after and also the z-axis torque monitor shows a monotonous and converging curve (see picture).
Could anyone explain me why there's this strange behaviour?

I also attach the picture of residuals.

Bye.