To add to this, I want to stress on the point of rarity. Variations within sex metrics are not the uncommon fringe case people make it out to be, they're actually common and expected.
Within biology, we'd see a number of metrics (like height) which would usually appear bimodal (like two bell curves added together). We might identify at least two latent variables here: A real-number 'age' (which can be observed) and a binary 'sex' (not directly observed). But it's worth stressing that these implied underlying curves overlap, and any given metric is not strictly correlated with the others. (Commonly, one might be on the lower end of some distributions and the higher ends of others. Someone can be 5'3" tall, have red hair, and a high body-fat percentage while also having testicles, XY chromosomes, and dying at the age of 62.) (We should also note that the 23rd chromosome just another observed variable, starting after ~1900.)
Some causes of variation that we know about are fraternal birth order, or endocrine-disrupting chemicals like PFAS, conditions like PCOS, etc.
Case in point are all the cis women who are impacted by the ever-stricter testosterone guidelines in the Olympics. Further is the effect of fraternal birth order, or the endocrine-disrupting chemicals like PFAS, or the intentional introduction of hormones and hormone blockers. (If certain industries are to be believed, soy milk has a similar effect.) These are all variations and things which impact what we understand as "biological sex".
Folk gender theorists tend to consider sexuality, identity, biology, and expression as orthogonal axes. But these are clearly also correlated among people. (Stretching the definition of "correlated" to include qualitative metrics like 'expression' using the usual methods.)
An information-theoretic framework would inform well an "optimal" way to talk about this, using a one-bit string for most people and increasingly more bits when more information is needed. This is roughly how people already talk.