I can't answer from the point of certainty of having looked at the code that runs WikiTree. I can't be certain because, contrary to my 3rd year vertebrate evolution course, there's more than one way to skin a cat. That's typical for computational challenges (cf. TMTOWTDI; or The Zen of Python)
Don't worry dear kitty - it's only a metaphor here!
So I can't say for certain how it's done, but I'll offer an educated guess and a counter-example to your line of thinking: The code determining one's connection might not be looking for a connection to a particular individual, but rather simply to the largest connected component of the graph. i.e. You do not need to have a "central" or "reference" individual.
But I should back up and explain what the heck I'm talking about...
WikiTree (and any family tree, for that matter) is a graph. Rather, WikiTree is a database of individuals and relationships, which can be described by a graph.
An undirected graph.
A family tree. Note: the depiction of nodes and edges is present different meanings when shown in a family tree format; it's purely a historical and stylistic choice.
WikiTree's family tree graph can be treated as being a directed graph or an undirected graph. The nodes (things that get connected; a.k.a. "vertices") are individuals; the edges (lines that connect the nodes) are relationships. A relationship, such as "father" or "mother" or "son" or "daughter" is a directional relationship, but we'll ignore that for now. I'm just telling you to give you the bigger picture.
A directed graph.
The relationships considered or included might depend on culture, e.g. in Western culture, we typically stick with relationships described by kinship terminology + marriage; compare with a genetic perspective, where marriage relationships don't exist separately from any offspring. This is the difference between having only one node - Prince Charles - between Queen Elizabeth II and Camilla, Duchess of Cornwall, versus having several nodes separating them. WikiTree makes the former choice, counting Charles to Camilla as an edge.
Numerous algorithms exists to analyze the connectivity of graphs and to determine whether nodes are connected or not. A graph may consist of several clusters of nodes that are mutually connected, but which aren't connected to all of the others. These sub-graphs are sometimes referred to as components (or subgroups).
An image showing the different subgroups or components detected in a graph data set.
Many network analysis tools include an algorithm to determine the largest component. Currently, if we were to analyze all of the profiles on WikiTree as a single graph, the "global tree" would be the largest component.
The largest component of various graphs as determined using the "Giant Component" function in NetworkX.
Hence it could simply be a matter of having the most connected individuals that determines what is the connected "global tree", rather than a connection to a particular individual. If that were the case, given the relatively small size of WikiTree's largest component, the connected portion of the "global tree" (13,763,773 people are connected at the time of writing), it would be entirely possible for some non-Western, mid-sized countries to overtake the current "global tree" that centres largely on western countries. (I, for one, welcome our new Elbonian cousins!) According to Wolfram Alpha, there are 24 African countries with enough current population to achieve this (without counting deceased ancestors, although they would certainly be necessary) and some of which might be "distant" enough that it would take a while before reconnecting with the main of WikiTree.
Of course these are just a few of the potential global rivals.
Although, I truly could imagine that some brave — and very dedicated— WikiTree Connector might undertake a strategic marriage if it could unite the two trees, should no alternative path could be found: