[ryanbr]

Issue #2566. WhoopRepository carried two functions computing the same value.

	hrUnionFingerprint/3 (line 1185, #2293)	hrFingerprintWindowUnion/3 (line 728, #2543)
signature	(activeDeviceId: String, from: Long, to: Long): String	identical
ids walked	rawWhoopSourceIds(activeDeviceId)	identical
per id	dao.countHrInWindow + dao.maxHrTsInWindow inline	the same two, via hrFingerprintWindow
per-id text	"$id=$count:$maxTs"	"$id:$count:$maxTs"
separator	,	|

I added the second one in #2543 without noticing the first.

Why the encodings were free to change

Neither string is observable. StressLensCache is an in-memory object holding one key, and PhysiologicalStepCycleEngine.loadCache is a plain HashMap field of a private data class. Nothing persists either value or sends it anywhere, so both doc comments were correct that the format is free, and switching a caller from one encoding to the other costs nothing beyond a cold cache at process start, which it always has.

Why it was worth removing anyway

The redundancy was not the risk. Two witnesses of one fact that are allowed to drift is the risk, and the fact they witness is whether cached scoring work can be reused. Teach one union to walk an extra id, or to carry a third aggregate, and the other keeps answering the older question silently.

The change

hrFingerprintWindowUnion is gone. The daytime stress lens memo reads hrUnionFingerprint, which now calls hrFingerprintWindow per id rather than repeating the two DAO reads inline. One union, one encoding, one per-id read. The cycle load cache keeps the exact value it had, since the survivor is the one it already used.

Strand/Data/Repository.swift named the deleted function as its Kotlin twin, so that pointer moves to the survivor. Both sides now also say what KIND of twin this is: each compares its own value against its own previous value, neither is persisted or transmitted, and the two encode the same facts differently. There is no byte-identity contract and no oracle asserting one. Without that note the next reader could take "twin" for the byte-level pair this repository usually means by the word, and align the encodings believing they were repairing something.

Tests

Neither union had a single test, which is how a duplicate survived review in the first place. HrUnionFingerprintTest pins the property that motivated a union at all (#908): a row landing under an ALIAS id, or under the canonical id, must move the witness although the active id is untouched. A fingerprint keyed on the active id alone passes a naive equality check and fails these. The stubbed DAO throws on any method other than the two index aggregates, so a row fetch creeping into the cheap path fails too.

Mutation-verified rather than assumed, with both mutations chosen to disable the guard rather than merely perturb it:

test	walk only the active id	drop maxTs from each entry
walksEveryUnionIdNotJustTheActive	RED	RED
movesWhenARowLandsUnderTheCanonicalIdOnly	RED	RED
emptyWindowIsStableZeroes	RED	RED
movesWhenARowLandsUnderAnAliasIdOnly	RED	green
canonicalActiveIsNotDuplicated	green	RED
stableWhenNothingChanged	green	green

Different kill sets, so the tests discriminate distinct properties rather than all keying on one thing. Every survivor is explainable: movesWhenARowLandsUnderAnAliasIdOnly also moves the count, so dropping maxTs still registers a change; canonicalActiveIsNotDuplicated sets the active id TO the canonical one, where the union and the single id genuinely coincide; stableWhenNothingChanged asserts determinism only, which both mutants preserve.

Verification

• Full Kotlin suite: 6528 tests, 0 failures, 0 errors, 768 classes. Confirmed actually executed rather than served up to date: every result XML carries an mtime inside the run window.
• parity_ratchet.py --base origin/main: errors=0, no warnings.
• Authority delta is exactly functions 4508 to 4507 and unpaired_functions 4109 to 4108, nothing else. Debt decreases by one, so no disposition is required.
• doc_comment_lint.py: green.
• Re-derived AFTER rebasing onto feat(oura): Experimental “Oura ring: all-day heart rate & HRV” — stand the daytime-HR hold down for the learned night, not for every screen-off #2558, not before. Deriving first would have produced a plausible-looking authority for a tree missing that PR's nine functions and two pairs, and the counts alone would not have shown it.

Note for #2567: neither of these functions can pair, because Strand/Data/Repository.swift is outside the ledger's Swift scope. That blind spot is how two Kotlin functions computing the same value both survived review, with nothing able to notice they were twins of each other.