User:Ovinus/GA investigations

Annoyances

 * Talk:Coral/GA2 – copy-paste job, not created by the actual reviewer. Multiple such reviews exist, transferred by (e.g., Talk:Somerset/GA1, Talk:Somerset/GA2 ...)
 * Solution(s): Ignore; compare user links in initial revision to creator; search for signature with lowest UTC timestamp and use that as reviewer; mark for manual review when the page creator does not match signature with lowest timestamp. (Secondary issue: renamed users. Possibly combine with renamed log, if dumps of that can be found.)
 * Going through the entire history dump (~300 TB unzipped) takes several hours. Random access is not possible because of the format.
 * Solution(s): Separate out page histories for GA reviews only; develop custom format for random access; pursue faster compression/decompression algorithm based on consecutive diffs.