Resolve "rank regression" when HT data unavailable for members in alternate sources

[tehhowch]

Some (all?) users with intermittent / "stale" data in Jack's FusionTable experienced a rank regression after the fix for HT 502's went in: Old values of LastSeen were submitted as new records and then propagated to the Scoreboard.

After HT came back online, visiting these profiles with the HT extension would result in the restoration of expected crowns & rank.

TODO:

• [x] Some, all, or subset? Interrogate the Rank DB to determine which members had the regression
• [x] Crosscheck with Crowns DB
• [x] Determine when the regression began (prelim estimate: Oct 14, 9:13 PM)
• [x] Why didn't this regression happen immediately when 502s were handled? What made the 14th special? Or did this happen immediately?
• [x] Devise plan to prevent repeat occurrence
• [x] Prune bad data

MAYBE:

• [x] Fix affected members' LastCrown calculated fields for relevant neighboring records.

[tehhowch]

Inspection of the Rank DB shows this has happened on more occasions than the 14-16 Oct, but went unnoticed.

• Some member records show deviations of ~15 seconds, so the overlap window for equivalence between HT and LatestCrowns should be increased, probably to 1 or 2 minutes.
• Earliest regression appears to have happened on June 9th (1528524810941)
• Inspecting the Crowns DB shows that some members have alternating data insertion between the data source values

[tehhowch]

Issue is suspected to be caused by the "courtesy insert" for members who have not had data inserted in the past 7 days. Will leave this open for the next week and re-run the regression-fixer then to determine if the issue is adequately addressed.

[tehhowch]

Re-ran regression fixer today, no detected regressions.