There's been talk in the past that smartpricing is no longer account-wide as it had been in the beginning. Recent data, however, doesn't support that idea.
The latest hit happened at the same time against more than one website under the same account. The second site hit is for a different industry than the first one, although it's the site under which the Adsense account was created. (The site hasn't been changed in over a year and it has the same SERPS, isolating key variables. Same ads, same placement, same colors, etc.)
So the smartpricing algo may review all sites under the same account when one or two hours of extreme data (or calculations) trigger an automated smartpricing penalty, and impose the same penalty on all your sites.
In the final hour before the day the smartpricing began, the CTR on one main page was at x%. Typically this number stays about the same one hour later (because it's an accumulation of data for 23/24 hours of the entire day, and traffic is at its lowest in the last hour of the day). The CTR percentage drop in the last hour is often less than 1% (x-1%).
But when the final stats for that day were reviewed the next morning, there was a 9% drop (x-9%). This suggests Google's data collection was delayed/inaccurate and in fact could have inadvertently caused the smartpricing.
There have been times in the past when Adsense users have noticed delays in stats, but is it possible some of the stats are delayed while others are on time?
(I've suspected on more than one occasion that Earnings have been accurate while all URL Channels were not completely updated.)
The next morning, as said already, there was a big jump with much higher CTR% (much more than 9%).
It's possible that, due to a glitch, Google mismatched data it collected for total impressions and total clicks over a two-hour period, so these two streams of data were out of sync, leading to an inaccurate conclusion.
For example, total impressions for one hour could have been matched against total clicks for two hours, instantly doubling the Click Through Rate. That could obviously trigger smart pricing.
In any case, it looks as if the smartpricing algo is much too sensitive. The previous week had CTR and eCPM the same every day +/- 1% (as had the previous months). An entire week or month should determine smartpricing, not one or two hours. Obviously two hours doesn't provide enough accurate and reliable data to draw a reasonable conclusion about site performance.