Also, these are subscriptions we're talking about. There's a reason the subscription model is such a powerful one. Without a subscription, people actively decide when they want to pay money for a product or service. With a subscription, the onus is on the individual to stop paying that money. They are much more reliable, even in the face of drops in functionality.
I'm not saying a) is certainly the truth, but it's not like it doesn't make any sense. You also need to avoid falling prey to the idea that there is widespread dissatisfaction with DDI. Some people are unhappy, but most of us think it's great, and we'll continue to subscribe for as long as we run 4e games.
Again, there's an easy way to check. Let a subscription lapse, and have someone check the group's members to see if they were removed.
The membership list shows 10 members per page, and 8125 pages. There are only nine members on the last page of the list. So that's 81249 members actually listed, and 81263 members claimed, a difference of 14 members.
If you were looking specifically at the Members tab, the discrepancy is probably due to the moderators and admins for the group not being listed in the Members tab.
There's a pretty straightforward way to test this. Find someone who had a subscription but let it lapse. Get his D&D Community username, then have someone with an active DDI subscription check whether that username is in the DDI group.
Also, notably, there was a point in the past where you were only added to the DDI group once you created a D&D Community account. I'm not sure whether the process is different now, but if it's the same then DDI group membership reflects only a fraction of actual subscribers.
Obviously it isn't possible to really say if the COUNTER for the number of group members is really accurate or not.
So... the two possibilities are
a) DDI has never suffered a noticeable net loss
b) the numbers in the DDI group are not accurately go up and down
Between those two possibilities, you're going with a). The number of DDI subscribers has only grown, despite the edition being dead with the last release being almost a year ago.
Despite the massive unhappiness in the fall of 2010 when the second annual payments were due and there were yet no replacement tools and everyone was upset about Essentials.
Every single month there was still more people currently subscribed to DDI than the month prior.
Or... the groups aren't registering people leaving the group properly.
I think I'll go with minor technical glitch.
As others have noted, it appears to be almost perfectly accurate. You can list the members of the group, 10 to a page. The number of pages lines up almost perfectly with the reported number of DDI group members divided by 10.