These items revisit the platform cases with new teams and new numbers. Each one asks for a decision that keeps a shared system healthy for the teams building on it: claim work once, buffer telemetry, map keys stably, overlap quorums and narrow work before expensive steps.
Before choosing, name the failure the decision prevents and who would notice it first. Platform failures tend to be quiet at first and loud later: a duplicated job, a label explosion, a cold cache after a resize or a replica that never caught up.
Apply the platform case decisions to new workloads and failures.