While doing some testing on the validation harness, I have run into a pretty trivially reproducible problem.
Despite two million rows being written, I have found that select * from standard1 only returns one page's worth of data. I have used both the java and the python driver to test this. I have also found that select count from standard1 gives a multiple of one page's worth of data, that appears to correspond to page size * RF.
I have already tried with the patch for
CASSANDRA-9775, and that did not resolve this issue.