CI Tests fail for some models

Some tests currently fail after the newest IBPSA merge, like here due to divation from reference results.

First @MichaMans and me thought this was due to the update from reference results in the IBPSA to dymola 2020

But even after using the dymola 2020 container the deviation changes but is still there, this can be seen here

I simulated one of the failing thermalzone examples locally, both with dymola 2020 and the deviation is still there, I used the same solver with same tolerances. The screenshot shows the timestamp with the highest deviation recording to the regression test.

@MichaMans @PMehrfeld do you have any idea where this deviation comes from?