Why is the choosing best Augustus species tie-break random?

I've just run funannotate v26.6.21 on a few sets of genomes where we have more than one strain per species, and looking at the logs for the `Choosing best augustus species based on taxonomy: ...` lines were all different. I traced this in the code to a random choice within the same taxonomic class, called via function `choose_best_augustus_species` here: https://github.com/nextgenusfs/funannotate2/blob/v26.6.21/funannotate2/utilities.py#L533 Why is this tie-break random? Are all the fungal Augustus models of similar quality, meaning you would expect much the same output from training from any given starting point? To minimise this potential source of variation between strains/isolates/species annotations, I should probably use `funannotate2 train --augustus-species ...` explicitly so that (for example) it is fixed to the same Augustus species for all my genomes of a given genus.