Really useful framing — the ledger idea forces the question of what each dollar is actually buying, not just what it costs. The point that stuck with me was the cut that looks safe on a spreadsheet but quietly removes recovery: running standby only on spot, or shrinking standby without a load test. A standby that cannot absorb real failover traffic is just a line item, not redundancy. The N/(N-1) capacity math is a good gut-check too; with two regions at 200% of peak, it is easy to see why teams drift into under-provisioning the second region. Thanks for spelling out the drill-after-every-cut rule — that is the part most cost reviews skip.