Our work is closer to tool and die makers, we make money printers that keep running without our day-to-day involvement.

Let's take S3 as an example. AWS S3 was started with 10 engineers or so (according to Andy). AWS S3 is now an organization with over 800 SDEs alone, that headcount costs well over $300MM per year (still excluding all the SDMs, Ops, QAs, PMs, TPMs, DCAs, etc also involved). That headcount made sense as the product exploded with new features, hardware, regions, etc etc... but the "greenfield" new feature development pace is likely an S-curve, and we are now on the slowing side that will only continue to slow.

How many SDE hours/year are needed for S3's steady state operation? How do companies compensate skill for rapid growth, and then transition to their steady-state needs? What changes in terms of headcount and/or compensation, if anything?

(Of course this isn't about S3, the same questions apply to any software product)

How would unions operate in this environment? It seems quite different than, eg Boeing's manufacturing unions with the long lead times and relatively stable production volume/labor needs.

(There are also interesting questions for shareholders - what happens if S3 lays off 90% of their SDEs? Where would those people go, what would they be most valuable working on - how "safe" is S3's revenue as competitors and startups hire their layoffs?)