kOmegaSSTDES model converges to unphysical result

I was using kOmegaSSTDES turbulence model to run a simulation at deltaT = 0.000002. The max Courant number is less than 0.01, but the result is clearly unphysical. I then decreased deltaT to 0.000001. The result is much better but the time cost is increased. I was looking for the reason so I changed the turbulence model to dynamicKEqn with deltaT = 0.000002. the converged result is also physical. It seems that there is a problem with the kOmegaSSTDES turbulence model, which will result in unphysical result unless we limit the deltaT less than 0.000001. I am wondering if there is something wrong with my case. My case files have been attached.SST_2D3.zip

There are three sets of result figures in my file. Result 1 corresponds to kOmegaSSTDES with deltaT = 0.000002, which is the unphysical result. Result 2 is kOmegaSSTDES with deltaT = 0.000001. Result 3 is dynamicKEqn with deltaT = 0.000002.

To reproduce result 1, you can use the following command decomposePar mpirun -np 8 pimpleFoam -parallel paraFoam -builtin

To reproduce result 2, please change the deltaT in controlDict to 0.000001, then decomposePar -force mpirun -np 8 pimpleFoam -parallel paraFoam -builtin

To reproduce result 3, please change the deltaT in controlDict to 0.000002, and change the turbulenceProperties to dynamicKEqn, then decomposePar -force mpirun -np 8 pimpleFoam -parallel paraFoam -builtin