There is an open PR by @ajwerner that would need dedicated attention to land. It (dramatically) improves the ease with which we can GC large version histories of a key.

The reason the PR is held up is uncertainty about the possible negative impact of many range deletion tombstones. This is a concern that needs to be investigated.

This issue tracks the future of that PR, a job previously done by (now closed) #50194.

#51230

Jira issue: CRDB-3015