Kevin R. Thornton

> would be great to side step that all using tskit stats.... Yes, the challenge there is that many of the stats are haplotype-y, and many such stats have an...

This looks like it is on the right track. One source of confusion, for me at least, is how msprime's "new" ploidy argument gets handled here? This came up in...

> But yes, I think it would be best if we did things in the general case with arbitrary ploidy. Yeah, seems prudent to iron out any wrinkles ASAP

I like the approach overall. I think embedding the `demes` bits is quite elegant.

Well, let me try to break a tie and see if I make a plot that looks more like @jeromekelleher or more like @mufernando.

```py import tskit import msprime ts = tskit.load("treefile_even_larger_effect_equilibrium_higmu.trees") for j in range(5): mts = msprime.sim_mutations(ts, rate=1e-9, random_seed=j + 1) ```  Tie goes to @mufernando, I think? Did this with...

For completeness, I repeated my example w/`gc.collect()` after each bout of mutating. 

The main difference between @jeromekelleher's example and the others seems to be the presence of metadata in the latter, which must be a clue?

> Just looking at your plot again @mufernando - is that really 20s to load a 250mb file? Damn that's slow! Hmm. Yeah. In mine, it is 14 sec for...

These experiments also gave a flat profile: ```py import msprime import numpy as np import tskit ts = msprime.sim_ancestry(10 ** 4, sequence_length=1000, random_seed=1234) ts = msprime.sim_mutations(ts, rate=1, random_seed=1) tables =...