Despite designing a row key with a UUID prefix to prevent hot-spotting in your Bigtable clusters, you're experiencing unexpected hot-spotting during data writes. What might be the underlying issue?