I take this as another sign favoring transcension over expansion, and also weird-universes.

The standard dev model is expansion - habitable planets lead to life leads to intelligence leads to tech civs which then expand outward.

If the standard model was correct, barring any wierd late filter, then the first civ to form in each galaxy would colonize the rest and thus preclude other civs from forming.

Given that the strong mediocrity principle holds - habitable planets are the norm, life is probably the norm, enormous expected number of bio worlds, etc, if the standard model is correct than most observers will find themselves on an unusually early planet - because the elder civs prevent late civs from forming.

But that isn't the case, so that model is wrong. In general it looks like a filter is hard to support, given how strongly all the evidence has lined up for mediocrity, and the inherent complexity penalty.

Transcension remains as a viable alternative. Instead of expanding outward, each civ progresses to a tech singularity and implodes inward, perhaps by creating new baby universes, and perhaps using that to alter the distribution over the multiverse, and thus gaining the ability to effectively alter physics (as current models of baby universe creation suggest the parent universe has some programming level control over the physics of the seed). This would allow exponential growth to continue, which is enormously better than expansion which only provides polynomial growth. So everyone does this if it's possible. Furthermore, if it's possible anywhere in the multiverse, then those pockets expand faster, and thus they was and will dominate everywhere. So if that's true the multiverse has/will be edited/restructured/shaped by (tiny, compressed, cold, invisible) gods.

Barring transcension wierdness, another possibility is that the multiverse is somehow anthropic tuned for about 1 civ per galaxy, and galaxy size is cotuned for this, as it provides a nice sized niche for evolution, similar to the effect of continents/island distributions on the earth scale. Of course, this still requires a filter, which has a high complexity penalty.