Non linear time dependent problem - Natural Convection

This is with respect to the code: Non linear time dependent problem - #2 by frederichecht

In this code, natural convection within sqaure cavity (bottom wall hot, top wall cold) is being simulated. For such flows, convection starts at Rayleigh number, Ra ~ 2585 (value obtained from literature).

As I understand the code, the first Rayleigh number to be used in this code is such that it gives a trivial solution (Ra = 1e3), i.e., there is no convection. The next Rayleigh numbers should be such that it gives non-trivial solutions. If I use Ra = 3000, it is giving wrong Nusselt number.

How much about the critical Rayleigh number should a Ra value be used to get correct Nusselt number results?