One thing that's not present here is support for resuming uploads: the file.allocate call will never match an existing upload based on content hash because we don't send a content hash (see D12065).

Instead, I'll match partial uploads based on file name, file size, author, and the "isPartial" flag in a future diff. I might add a dialog prompting users to resume, but <name + size> probably uniquely identifies every large file a reasonable user is likely to upload.