I did a comparison on your data sets such as DS1, and S3-FIFO had very lower hit ratios than LRU in most cases. I think there are very limited situations where S3-FIFO is best.
https://github.com/Thesys-lab/sosp23-s3fifo?tab=readme-ov-fi...
Disclaimer: This is the author of S3-FIFO.