Nope that's wrong, I had an account which was inactive for 3 months and it got hit, so yeah the detection was way further back than weeks.
Actually Ahemoth is correct!
Blizzard's usual banwave pattern is after the initial detection, they collect data for one to 3-4 weeks, then examine the records for a week or two and initiate the actual banwave.
Meanwhile they pull with the banwave-flagged accounts, these, flagged by different criteria in the same or close time-window too: Like player reports, heuristics flagged etc., which is easily confirmed just by watching the Ban section here:
We had suspicious low activity on bans, reported in the months before the banwave:
After the ban report spree from Dec 2015 banwave,
In Jan 2016 we had 46 reports.
In Feb 2016 we had 66 reports, ~50 of them just on 05.02.2016 after some large batch of flagged accounts got banned.
In Mar 2016 we had 7 reports
only! Quite a few, considering the billions-worth corporation Blizzard have a huge bot-hunting team, they delivered
just 7 ban-reports from 27.02 to 16.04 - in 49 days timewindow! If that is not suspicious low, what else?
In Apr 2016 we had 17 reports! Again quite low number!
In May 2016 we had 3 reports prior 19.06 banwave!
In May 2016 we had
192 reports from the banwave.
The pattern is more than obvious:
Blizzard were collecting a large batch of account flags for a banwave since late Feb 2016, nearly full 3 months prior 19 May!
So no surprise that they have banned and accounts, inactive from two-three months - they were simply flagged in Feb/Mar and banned with the wave.