Here is what this volume will tell you, without hedging, to the unit:
Seven days is how long you have and it is enough. A foetus is a bubble at five nights, a jujube-sized lump at ten days, has a head in a month, feels hunger in the fifth, and remembers a hundred previous lives in the seventh. There are three doors out of the body and only one of them is the good one. Fourteen divinities enter the cosmic body's organs one at a time, and the refrain "but Virāṭ did not wake up" repeats fourteen times until the right one arrives. The transit to Yama's house is ninety-nine thousand yojanas and takes two or three muhūrtas. Abhijit is the window to kill in, and even the disagreement about it is precise — the text says two muhūrtas, the tradition says one, and the endnote records the gap.
Now here is the same volume on what all of that is for.
Seven sheaths, or eight. The cosmic body mapped three incompatible ways, with 2(5) admitting a fourth exists somewhere. Two avatāra catalogues, differently constituted and differently ordered. Twenty-four-and-one tattvas in Skandha 3 that do not match the 4+16+5 scheme in Skandha 2. Four kinds of primal ignorance in one chapter and five, with different names, in another.
The road is surveyed to the yojana. Nobody agrees what is at the end of it.
[RESONANCE] Because I had the relationship backwards, and I think most people do.
I assumed the metaphysics was the settled part and the practice was the improvised part — that a tradition first works out what reality is and then argues for centuries about what to do about it. This volume is the reverse. The procedures are stable enough to time with a water clock. The ontology is a set of competing drafts kept in the same binding.
And once you see it you cannot unsee where the confidence lives. Confidence is highest in the sentences that tell you what to do next, and it degrades steadily as the sentences get closer to describing what is actually there.
There is a functional reason for this and it is not embarrassing to the text.
A procedure is testable by the person running it and an ontology is not. Somebody counted the weeks of a pregnancy. Somebody sat with a dying man for seven days. Those claims got corrected by contact with the world, over and over, until they hardened. Nothing corrects a sheath-count. So the parts of the corpus that were exposed to feedback converged, and the parts that were not stayed plural — and the plurality is not a failure of rigour but a record of which claims had a way of being wrong.
Which flips the vault's usual reading of the seven internal disagreements. They are not the text being sloppy about its own doctrine. They are a map of where the text had nothing to check itself against.
And then the translator glosses each occurrence to fit its local number, which erases exactly that signal. A reader with the notes and without the census sees a consistent book. [n≥3 PATTERN] on the local glossing; see Seven Things This Text Says Twice and Differently.
Against the vault's own habit of citing doctrinal schemes as though they were the tradition's position. If precision tracks testability rather than importance, then a numbered scheme in a Purāṇa carries roughly the epistemic weight of a house style, and the vault has pages that lean on those numbers.
Promotion criterion: run the same split on a second corpus — the Śiva Purāṇa build's material is on disk and already has an endnote census. If procedures there are also stable while ontologies vary, [n=2] and worth a real page. If the Śiva Purāṇa's procedures vary too, the reading is dead and the tidiness here is just this compiler's temperament.
Pick five procedural claims and five ontological claims from the Śiva Purāṇa corpus and check variance across its three volumes. It is an afternoon, and it either kills this or gives it an [n=2].