For .isel(), I think the natural behavior would be to drop any datasets that don't have dimensions that match the isel arguments. I don't know what to do about datasets that have those dimensions but the isel() values exceed the extent. E.g. a collection with ds1 and ds2, where both have dims named lat and lon, but ds1 is 360x180 and ds2 is 50x50. Should collection.isel(lat=100) return a collection only containing ds1.isel(lat=100) and maybe a warning that ds2 could not be included?