Comprehensive assessment of climate datasets is important for communicating to stakeholders model projections and associated uncertainties. Uncertainties can arise not only from assumptions and biases within the model but also from external factors such as computational constraint and data processing. To understand sources of uncertainties in global variable-resolution (VR) dynamical downscaling, we produced climate datasets using the Model for Prediction Across Scales dynamical core coupled to the Community Atmosphere Model version 5.4 (CAM-MPAS). CAM-MPAS is configured with VR meshes featuring higher resolutions over North America (50-200, 25-100, and 12-46 km) and run for both the present-day (1990-2010) and future (2080-2100) periods using pseudo-warming experiments. The experiments use sea surface temperature (SST) and sea ice changes from a global climate model in the Coupled Model Intercomparison Project (CMIP). Resolution-sensitivity of the hydrological cycles appears consistently in the present-day and future climate in CAM-MPAS, most strongly over the tropics, while the mid-latitude circulations in the Northern Hemisphere are not very sensitive to resolution. The SST constraint leads to a similar future precipitation change over tropical oceans in the CMIP model and VR CAM-MPAS. Over the land and sea ice, projections of near-surface climate diverge between the two models, as do the upper-level circulations across the globe. Continuing spin-up trends in deep soil are found over the high-latitudes, which may affect the local surface energy balance and remote circulations. Precipitation statistics improve with higher resolutions within the refined domain, and such statistical inference is verified to be uninfluenced by horizontal remapping during post-processing. Such global and technical aspects of VR downscaling require further investigations to reduce uncertainties for a target climate region.