With this change, we started seeing OutOfMemory issues on the server. It seems that for large file uploads this would allocate a massive buffer in memory (only in cases where the content type isn't specified?). The buffer should only need to be a portion of the overall content length since it parses in chunks at a time.

When patching to restore the old behavior (let Werkzeug set the default buffer size) then it resolves the memory issue and we don't experience issues with parsing really large requests.