While I am very pleased with VEB overall, I have seen a fair number of backup chain corruption events, where the fix was to wipe the database and start a new chain. The most recent was Case # 01907423. My feature request is to make the backup chain more resistant to corruption.
The worst thing about chain corruption is, no possibility of restoring a system if it fails before the chain is restarted. The next worst thing is, even after the chain has been restarted, the old data is not recoverable.
I think this could be improved by adding changed blocks to the VBK during the merge process, but not committing the changes and removing the aged out blocks until the merge has completed. So if a merge fails, you're left with a larger VBK, but all the pre-merge data is still there and the VBK is still usable for recovery.