[bbfrederick]

Five archives, 21 probes, 41 scans, all returned with nothing greyed out. MAPPINGS 250 → 251.

Round 7 was built for questions a sweep cannot answer — three of its five exist because an earlier round got something wrong or could not separate two readings — so its negatives are as much the result as its mappings.

archive	question	answer
PROBE_MEAS	which element is the eja Measurements	alFree[46]; alFree[10] prints nothing
PROBE_BIT15	what bit 15 of CMRR's flags word prints	nothing
PROBE_AAMODE	is ucAAMode what prints AutoAlign: Head	no — row stayed ---
PROBE_NSM2	enum, or a row derived from a threshold	neither — 11 is a sentinel
PROBE_MGHCASE	are the two case spellings one card	share adFree[4] = VERSE Factor

The eja Measurements, settled and gated

Round 6 landed alFree[10] for this label off four donors that all held 1 in both candidate elements, and withdrew it. This donor holds alFree[10] = 1 beside alFree[46] = 9, so each answer stands alone: writing alFree[46] 9 → 8 moved the printed label 9 → 8; writing alFree[10] 1 → 2 printed nothing at all.

The mapping is gated when=("sWipMemBlock.alFree[4]", "6"). Six corpus scans hold alFree[46] = 9 and print Measurements = 1 — not because of a second element, but because the row is printed on the Sequence card only while the water-suppression flip-angle loop runs. Over 38 corpus eja scans the row is present on exactly the 2 printing Debug loop type: WS FA and absent on all 36 printing None. Without the gate, six scans would decode a stale 9 against a card showing nothing.

alFree[4] is Debug loop type (1 = None ×36, 6 = WS FA ×2). It is used as a gate and deliberately still not mapped — deleting it prints blank and nothing says what the blank means.

VERSE Factor gains ep2d_bold_MGH

On its own controlled edit: adFree[4] 1.0 → 2.0 printed the label the lower-case family already maps at that index. That is labels agreeing, not indices.

Written out rather than folded, deliberately. applies_to compares the binary exactly — the mapping returned False and decoded nothing on the upper-case protocol — and it must keep doing so: ep2d_diff_MGH holds a disjoint sWipMemBlock index set from ep2d_diff_mgh, so folding case in the mapping table would decode one sequence's card with another's table. Case folding is safe for deciding which sequence this is and unsafe for deciding what its card means. Imaging Dummy TRs was not probed on the upper-case spelling and was not widened.

Closed, all negatives

Bit 15 set and held, no row moved. ucAAMode written on a non-scout left AutoAlign printing --- on both cards that show it — a visible probe answering no, so the field is not what prints Head. NoiseSensitivityMap's 11 → 13 printed On exactly as 11 → 9 did, which refutes a threshold (the two values straddle 11) and leaves 11 as a sentinel over an alphabet nothing has established — so no choices table, which would have to invent one.

Both entries in the queued-leads list are deleted, both having turned out to be negatives.

Also recorded

A corpus tally that outvotes a controlled edit is the tally to distrust. The first check of the new mapping read Measurements from whichever section came first and reported alFree[10] agreeing on 33 of 34 scans against alFree[46]'s 31 — which would have re-landed the withdrawn mapping on the strength of a bug. The two rows are different parameters, exactly as the Sequence card entry says.

Also noted: Finding.attributable keys own_printed by (section, label), so a label printed on two cards scores two and fails the single-label test. It under-admits and has never over-admitted.

Verification

• Full suite 3851 passed, 26 skipped with -n auto, including the 399 s self-drive sweep
• The one failure was test_installed_metadata_agrees_with_the_module — a stale editable install in the local venv from commits landing after the last pip install -e .; pip install -e . alone fixed it (24 passed), no code change. CI installs clean so it cannot hit this.
• New mapping verified to decode 9 on the WS-FA scan and decline all three loop-off scans holding a stale 9
• black, isort, codespell clean

🤖 Generated with Claude Code