Yeah was thinking of curse of dimensionality, with standard Euclidean distance for example, distances in high dimensions lose their meaning, but there are surely other distance which could function there. Btw I have used sklearn to cluster on couple millions of rows with sklearn successfully