Complex file renames
I've fixed the server algorithm that pulls together pieces of metadata from renamed files into a single synthetic "master" file. While it always worked fine in simple situations, it used to get weird (and outright wrong) results for more complex rename graphs, such as when you renamed A to B then later C to A. The algorithm for determining whether a rename occurred remains unchanged.
This may affect the review state presented to both the built-in review completion computation and custom review completion conditions, and hence also the output of these conditions. Please let me know if you spot anything untoward, especially when it comes to custom per-file reviewed flags since Reviewable needs to map those from the synthetic file back to its constituent parts.
Note that the file synthesis algorithm on the client remains the same for now, and I think doesn't suffer from the same bugs, but will be getting updated to the shiny improved one in the next major release anyway.