This looks to me more like a secondary index issue. If you say the access
via rowkey is always correct, then the repair works fine. I think there
might be something wrong with your secondary index then.
Just a follow up in case someone will have the same case:
This problem was solved by running major compaction on Cassandra's
index CF (the internal one, not the indexed one) via JMX. After this
there was one more user that was gone in index queries (no matter if
CL.ONE or CL.ALL), but I've rebuilt the index and it was solved too.
Still don't really know what happened (looks like something related to
indexes and their tombstones or so), but it's fine now.
M.
