# Reclassify What Blocks Before Optimizing What Blocks ## Claim When a process is failing under load, changing which steps are mandatory-and-blocking versus scheduled-and-advisory produces a larger and faster win than optimizing the mandatory steps further, because it removes work from the bottleneck instead of speeding it up. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Requires knowing which checks actually gate a decision; a check nobody acts on is advisory already and only nominally blocking. - The risk is real: reclassifying a check that does catch regressions trades latency for escaped defects, so it needs the catch rate measured first. - Bounded by checks that must block — a security gate or a release verification is not a candidate. - Same reasoning as moving work off any critical path, which is why the win is usually larger than tuning predicts. ## Evidence From the LeadDev piece on halving a CI pipeline through sharding and flake triage, where reclassification outperformed further optimization. ## Links - [[A Capacity Fix Sized to One Volume Driver Expires When the Driver Changes]] — the other lever from the same piece, on sizing rather than sequencing - [[Managers Should Code Off the Critical Path]] — the same principle applied to where a person's work sits rather than a pipeline step - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — the amplifier thread on relieving a constraint AI made worse. - [[3. Resources/Permanent Notes/Running Work Less Often Can Raise the Worst Frame When Every Instance Shares One Schedule]] — applied in: Cutting how often a repeated job runs lowers total work, but if every instance moves to the same new schedule, the instances fire together and one frame takes the whole load.