Spot VM lifecycle — single controller and kill switch pattern
Maintained by PointSav Digital Systems. Each revision is content-addressed by its commit hash.
- Track-B documentation wave: infrastructure category — 5 severe R1 infrastructure-exposure redactions (internal paths, commit hash, hardware IDs, tenant IP:port pairs, real costs), 1 entity-capacity fix, 1 EN/ES parity restoration
- editorial(infrastructure): fix spot-vm-lifecycle-kill-switch throughout + add missing index_group (Track-B) — de-narrated a standing correction block and applied its findings consistently: real orchestrator is nightly-run.sh + nightly-run.timer (not yoyo-daily-cycle.sh/local-yoyo-daily.timer), real idle monitor is an in-process Doorman background task issuing instances.delete (not an external yoyo-idle-monitor.timer), kill-switch mechanism itself confirmed accurate; resolved the standing open question on training-trigger.timer by reading nightly-run.sh directly — it already runs training every night as Phase 2, making the separate weekly timer file read as superseded documentation rather than a live second start path; fixed a pre-existing content-contract ERROR (missing index_group, never set on this article) to compute-and-vm-fabric matching the category's Index Topic; register-clean EN+ES
- correct(infrastructure): flag script/timer name drift, in-process idle monitor, training-trigger.timer conflict (spot-vm-lifecycle-kill-switch)
- editorial(titles): sentence-case re-casing across 43 documentation titles
- Phase C A1b: rebalance PPN/OS/substrate/services out of architecture/ (16 articles, 32 files); resolve 8 of 9 slug collisions (aliases + redirects); language-protocol-substrate left for editorial call