As far as I know it is a fundamental design difference between btrfs and ZFS. Btrfs has extent back references. Given a location on disk btrfs has an efficient data structure that gets from the disk location to the items using that disk location. E.g. if a sector on disk gets corrupted it has a command that lets you list all the the files using that sector. With ZFS you'd have to iterate through all files, list block locations and check if one of them is affected. Of course such an additional index comes with a performance cost for managing that index (any kind of write or delete operation plus more metadata space used). On the top of my head this is used for rebalance, quota groups, device removal, send/receive (finding reflinks) and file system shrinking.
For a similar trade off compare hard links on Windows vs. ext4 and co. on Linux. On Linux there is no command to list all hard links to a file while Windows (NTFS) has one (FindFirstFileName etc.). This is why creating hard links on Windows is much slower and there is a limit of 1024 hard links to a file. To find all hard links to a file on Linux one would need to iterate over all files and check if they have the same inode number, however.