Some of the problems in this thread are unfortunately inherent to the use of btrfs. For example, output from the system commands df and du will generally be different from that of btrfsprogs btrfs filesystem df/du. The reason is that filesystem metadata, RAID levels and subvolumes all make the concept of “free space” much more complicated, also depending on whether you look at a whole filesystem or an individual share. The developers themselves explain this better in their FAQ section.
There is another issue with “zero usage” showing after making changes to a btrfs pool; this particular one can be resolved by performing btrfs quota rescan /path/to/pool.
At the moment, the usage reported for pools should be the most accurate. I’ve started an attempt to get a bit more sense into the share usage as well, with some changes to the shares list view in the most recent (3.8-14.10) version. Notably though, I haven’t touched the dashboard view yet. I’m keeping a list of current problems in this github issue:
Feel free to comment or contribute there as well.