Little capacity for code sharing and lack of reasonable tools to do so. That and the fact that newer statistical / ML techniques are extremely expensive mean that if you wanted to do e.g. a transformer you'd have to either pay out the nose or roll your own from scratch.

That and my department of about a dozen can't justify the cost to upgrade to their more enterprise focused, cloud ready new version Viya, but there's not really an upgrade path for 9.4 other than minor maintenance releases. Since no young new hires come in with SAS experience and the whole thing seems like a ticking clock, we decided just supporting Python for new development is a better plan.