The time evolution of tropical hydrological cycles is important in terms of water resources for a large population and as a main driver for the general circulations, but is challenging to predict due to the wide range of scales involved, from individual convection to the Madden-Julian Oscillation (MJO). As such, a global model with convection-permitting resolution is a suitable approach, although it still relies on parameterizations of cloud physics, radiation, turbulent mixing and shallow clouds at unresolved scales. Previous work showed that adjusting tunable parameters can significantly affect the MJO representation in the Nonhydrostatic Icosahedral Atmosphere Model (NICAM). This work extends the NICAM parameter-sweep experiment to 1) another model - the Model for Prediction Across Scales (MPAS)-Atmosphere, here employing a 9-36 km variable-resolution mesh, and 2) including objective tracking of Mesoscale Convective Systems (MCSs). During an MJO event in December 2018, most of the eastward-propagating precipitation is found to originate from MCSs. A three-week long initialized MPAS simulation, using the physics suite for mesoscale resolutions, does not produce MCSs in the Indian Ocean. The convection-permitting suite with the Grell-Freitas convection and Thompson microphysics is able to produce MCSs over the Indian Ocean, but the precipitation intensity and number of MCSs are largely underestimated. Tuning a microphysics parameter to delay the ice-to-snow autoconversion rate further improves MCS numbers in MPAS. However, most of them do not propagate to the east, producing little MJO signal. We will test other parameters for their effects on MJO signals with MPAS and track MCSs in the previous NICAM simulations to ask if such a model tuning simultaneously improves the smallest (MCS) and largest (MJO) scales of organized propagating cloud systems in the tropics, and explore sources for improvements and their pathways across scales in the two models.