Platform SLO
A published reliability and performance objective the platform commits to for the teams that depend on it.
A platform without published objectives asks every consuming team to design against an unknown, and prudent teams respond by building their own fallbacks — which destroys the consolidation the platform existed to achieve.
The objectives that matter are not only uptime. Pipeline availability and p95 build queue time, provisioning API availability and time to provision, control-plane availability, telemetry ingestion delay, and time to restore after a platform incident. Each is something a consuming team plans around.
The uncomfortable arithmetic is dependency stacking: a product service targeting 99.9% availability that hard-depends on a platform component offering 99.5% cannot meet its own target, and no amount of effort inside the service will fix it. Publishing the number is what makes that conversation possible before the design is built rather than after the first incident review.
It also disciplines the platform team, which now has an error budget of its own and a defensible basis for refusing a change that would spend it.