When creating the virtualizarr, we used open_virtual_dataset and then virtual_ds = xr.combine_by_coords(virtual_datasets, coords='minimal', compat='override', combine_attrs='override') but when we created the virualizarr for 2019 to combine with the previous data, we used virtual_ds_2019 = xr.merge(virtual_datasets_2019, compat='override', combine_attrs='override') - why the different methods?