Naively, say there are three possible timescales for humanity, and we assign equal (33.3%) credence to each of them:
1. Short: Humanity dies out within 100 years or fewer
2. Medium: Humanity dies out within 1,000 years or fewer
3. Long: Humanity dies out within 1,000,000 years or more
In this case, the overwhelming moral importance still lies in the far-future (1,000,000+ years). So long as you accept the basic Atemporal argument of Attachments, the mere possibility of a far-future dominates the expected value calculus.
You could tweak the probabilities to assign 99% credence to the medium-term view and only 1% to the long-term view, and the math will still work out.
Growth will still matter in that it accelerates our arrival at the “saturation” point, but as estimated by Nick Bostrom in Astronomical Waste, the cost of this delay is miniscule compared to the cost of outright extinction. So existential-risk remains of tremendous importance, but where does that leave progress?
There is much more at the link. And here is the blog of Applied Divinity Studies.