[en] Monte Carlo eigenvalue simulation of large systems is prone to slow convergence which makes the diagnostics very challenging and the estimate of the statistical variance questionable. This paper investigated the entropy-based diagnostics in MCNP5 using the OECD/NEA fuel storage pool problem. The results indicated the improvement of non-convergence detection, but the slow converging problem remained challenging and non-convergence was not always detected. As one possible approach, the entropy bounding was examined. To further understand the convergence process, we analyzed propagation of the flux distribution through the problem. It turned out to be a potential robust convergence diagnostics approach in the new future. (authors)