Oliver, by chance are you also doing a TTL on the data? Or maybe you've already issued a DELETE that has a future timestamp?
The only other way I can see this happening is when the node's clocks are skewed. Have you checked whether NTP hasn't drifted by a massive amount? Cheers! GOT QUESTIONS? Apache Cassandra experts from the community and DataStax have answers! Share your expertise on https://community.datastax.com/.