At the very least see the amount of valid records on the RS pages increase, Zika at least per my hunting tool counters, something observed at times during weekends, but east of the Atlantic nobody notices as stats are then well ready... after shuteye. ;O)

BTW, from few nightly observations, MC lock-out does not start until about ~00:30 UTC, so this access did not get back until ~3:00 UTC?

No harm done... all suggestions welcomed.

Interestingly enough, the "extended" run time only started recently (maybe within the last week). Prior to that, the run time was relatively static. In other words, didn't organically grow longer. Last night was the first time I noticed that it went beyond 0200 UTC (by 5 minutes). So, it's still growing slightly.

Additionally, the 1200 UTC stats are also running slightly longer but only by about 10 to 15 minutes.

I noticed that the statistics were updated at 13:05 (the values had changed) but the time stamp still indicated that the statistics hadn't been updated in over 13 hours. Evidently there is a delay between posting the values on the MC page and actually updating the timestamp...

The stats process is due for an overhaul (everything from the process to do the data updates to what data we store). I don't think we have substantially altered it in 6-8 years and even then it was just performance improvements on what was a process that was developed back in 2004. I'd like to move us to a point where it is more continuous rather than every 12 hours (note that when I say things like this, the list of things I'd like to overhaul is much larger than my available time

Here is some info about the stats (from our logs):

**********************************************************
Fri Sep 16 00:06:01 UTC 2016 - Start
Fri Sep 16 01:23:14 UTC 2016 - Going Unavailable
Fri Sep 16 02:01:24 UTC 2016 - Making App Available
Fri Sep 16 02:04:55 UTC 2016 - Back Available
Fri Sep 16 02:04:55 UTC 2016 - Done
**********************************************************

Although the process took about 2 hours, the website was unavailable from 1:23 until 2:01 so about 38 minutes.

The longest step in the whole process (it took 61 minutes) was calculating the "user project stats". This is the table that stores the aggregate total by project by person. So we would likely start here to improve the total duration of the update. However, this step occurs while the website is operational so it doesn't really impact anything.

As far as the slowest part of the update when the website is disabled, it is the work to create the total stats for each team. This took about 15 minutes of the time that it was unavailable.

Ran the MC pull manually at 15:10 CET (13:10 UTC) and data had refreshed, but not the time stamp update, i.e. nothing has changed from previous [noon] observations.

As to

Fri Sep 16 01:23:14 UTC 2016 - Going Unavailable
Fri Sep 16 02:01:24 UTC 2016 - Making App Available

Distinctly remember to have seen very recently that the 'unavailable' screen indicated something like 43 minutes, when doing a latenighter coding/testing session. Maybe Doneske has more precise notations, but 38 minutes is not what was the experience that night... still lots better than being blocked from 00:06 start to completion.