Just post timing happenstance, Serena; sorry. I wasn't replying to your note.
It was just a meant-to-be-humorous, tongue-in-cheek comment about the Question title since the vast majority of those 5 million profiles aren't connected by any DNA evidence, only by test-takers entering on their profiles the fact that they themselves have taken a DNA test. That fact propagates all over the tree (per established criteria) and appears under the heading "DNA Connections" on all other possibly-pertinent profiles...and those entries are truly "hints" or "research suggestions," not "connections."
I'd love for WT to post the number of profiles that contain information about a DNA test being taken. Would be a much smaller number, not nearly as impressive, but that's the number that's most important. It's only the test takers that we can compare.
Counting "connected" profiles, a single error for a recent generation in the tree could propagate as dozens, perhaps scores, of erroneous autosomal "DNA Connections" hints. And even a lowly HVR1 mtDNA test--essentially of no genealogical use except as negative evidence to rule out a hypothesis--will propagate to matrilineal profiles as far back in time as any profiles exist.
A "Confirmed with DNA" profile count wouldn't really help, either, because I'd hazard at least half of all the ones on the tree either don't follow the WikiTree confirmation guidelines for evidence citation, or use over-simplified autosomal triangulation to distant cousins that can't be defended (from an evidentiary standpoint) and are probably incorrect.
Not being a naysayer; 5 million is a fantastic number. I'd just love to see both statistics: total "connected" profiles, and total number of WikiTreers who have taken a DNA test.