This approach doesn't work with DLT, because the manual DELETE command immediately causes pipeline corruption. We have to use skipChangeCommits=True, but this option has several limitations, at least it cannot be used with apply_changes(). Is there any other approach to cleaning up old data in DLT?