Humans gone through many thousands of generations of stable or decreasing population growth, including at least one serious bottleneck event.
In these generations, each woman left on average one reproductive female child - 25% 2, 25% 0. In the solitary female child one of her two grandmothers contributes no mitochondrial DNA - she got one of the X strands and the egg mitochondria from a maternal grandmother, one of the X only from a paternal grandmother.
So tracing any given maternal lineage, we see in each independent - unduplicated - generation a 25% chance of losing a given maternal ancestor from that line.
But as some disappear, duplicates of others - women with two daughters - fill in. Now the odds are different - extinction is not as likely with a larger population.
And so it goes, with early advantage likely to be reinforced and amplified. In the end, those choosing mates are most likely to choose from one of the duplicated mitochondrial lines (there are many more individuals in them), and if duplicated themselves are now certain to redouble the duplications.
After a while of this there are likely to be only one or two truly numerous lines,
and then come the bottleneck events, or the migration, or the geological isolation vents, or a disease, or the like.