My last census upload yesterday ended up with something I have never seen before. The New characters found [10376] was 40% larger then the updated character count [6185]!
One factor that might be in play is that yesterday I had very poor connection to battle.net. It would be fine and then the world latency would go out of sight leading to my client not being able to log out characters.
Closing the WoW client does correctly force the saving of Addon data, so character data was saved including activity and realm logging if you exit the client after a census run.
I had to restart census runs multiple time when the connection got so bad that I was just dropped completely and had to log back into battle.net.
All the submissions except the last one for yesterday were due the fact that I had been disconnected and had a convenient stopping point for uploading.
This is pure speculation... but you said you suffered from disconnections... Did you by any chance census any multi-connected realms? If yes... you might take a look inside your .lua and see if you have large number of characters that have been disassociated with their actual realms and now read as new ones on a realm they are not really on?
As another idea... was there perhaps a database cleanup? Large number of new ones might be a result of cleanup removing lots of "dead weight" and you happening to readd it?
Third option would be that you happened to hit a lot realms that just zeroed out and you picked up results from them. I seem to recall that in the distant past I sometimes got "abnormally" high new ones results.
We might get wiser if someone else also reports abnormal distribution of results... My own patterns seem perfectly normal and similar to how things were for example a week ago. Roughly one new for each about 5 updates.
I ran into the same problems yesterday. Had to end task WoW a few times because nothing else could get me out. Was a bit irritated as one of the WoW lockups/end task corrupted something and made WoW crash a few seconds after connecting every time. My numbers from yesterday are about as expected I guess. I hit a bunch of connected realm sets so some were closer to or a little over 50% new.
Good thought.. but I don't do CR's any more... they have plenty of census runs.. I'm working on realms that have less then 10 census total when you add both sides up.
After further thought, nope that can't be a reason.. CensusPlus doesn't work that way.. the only way for that to happen is for Blizzard to lose character-realm associations. That would be a game stopping disaster requiring server shutdowns and rollback of databases to a previous known good point. I don't think that has happened since early beta days.
As far as database clean up... I don't think Rollie or Metalbeast purge data at all.. they can and have put limiters on search and display selectors to speed those functions off and on.
The third option is possible... I only know a very very little bit about the db schema and the php coding... while the 30 day activity counts can go down. I think a new character is strictly that.. a new character, never seen before in that configuration... realm/character_name.
Any chance any of your realms had a recent (or since you last censused it) incoming free migration? That'd be the most obvious reason for more new than updated characters to me; low pop realm with not many characters suddenly gets influx of transfers and maybe a bunch of new-rolled characters too (maybe some server had an event with a zillion lvl 1s, although I'm not sure if <10 even counts towards new/updated).
That said, I've never managed more new ones than updated in any submission that I'm aware of, and I seem to have a pretty high number of new ones due to doing so many 0 census realms (far more likely to hit characters nobody else has hit yet on a realm that gets done once a month!), so your results are pretty bizarre there.
As I said, I was just speculating... However, I can now confirm that this is not entirely unique incident. User "yushiro15" is on the top tens... and the only submission from this year has far more new ones than updates. That submission was also made during a very similar time to yours. Could this be a server side error of some sort? Or something in general relating to the change of year? Or some coding, which only executes at some specific trigger point? Like having too many level 1s in the database?
There are some details that I can't quite recall, because they are not discussed very often... Are all cleanups manual? Does some data "auto-expire", like very old 1s or unused level 55-58 DKs?
Also, Balgair's point about free migrations is a very good one. Maybe one or more of your census targets recently had a large, so far undocumented, influx of characters after it had been at or near 0 submissions for a fairly long time?
stopping point for uploading.