Hi,
We have an issue opened for implementing tombstone scanned histogram,
which can be very useful for increasing observability of tombstone
impact on reads.
https://github.com/scylladb/scylla/issues/7636
As for tombstone_warn_threshold, we're still thinking about how we're
going to implement it. Bloating the log with warnings that are hard to
spot may be suboptimal. In ScyllaDB, we keep track of large
partitions, larger than a threshold, in a system table, and a similar
approach could be employed for tracking reads which # of tombstones
crossed a threshold.