Before the head unit is ever powered, both PSUs of every expansion shelf are cabled, with each shelf's two feeds landing on different PDUs. Why is this a precondition of head bring-up rather than later cleanup?

The head enumerates its storage chain at boot, so every shelf must be powered and settled — both PSUs cabled, with feeds split across separate PDUs — before the head wakes and takes inventory of the enclosure chain. A shelf that is dark at that moment is absent from the enumeration the bring-up depends on, while a single-fed shelf enumerates normally but joins the configuration carrying a redundancy gap that stays invisible until the first PDU event removes its only live feed. The task order exists so the first enumeration, and the configuration built on it, reflects a complete and fully redundant chain rather than one that has to be corrected later with a shelf rescan. Shelf inrush current tripping the head's feed fails as invented reasoning: distribution is sized per the design, and sequencing exists to establish enumeration order and complete paths, not to dodge startup draw. Gating enumeration on a license-record serial scan fails by concept: chain presence is an electrical and physical fact, and license state has no bearing on whether an enclosure answers discovery. Limited-cycle power cords fail as nonsense — they are standard rated assemblies intended for repeated service. Exam caveat: the pre-power checklist puts both shelf feeds live and separated on different PDUs before the head powers on. Operational check: verify each shelf's two PSUs are in live outlets on opposite PDUs, with normal PSU status, before touching the head's power.