Slots 000 were RECONSTRUCTED, not read cleanly.

The captures held the right content but were missing
bits, so everything after the gap sat shifted.  The
missing bits were reinserted inside a uniform fill run
(where the insertion point does not affect the result)
until the frame tail matched the format template
exactly, then the FCS was solved for.

Confidence: the text region is byte-identical across
several independent transfers, and 22 bytes of known
tail structure land exactly.  The FCS was computed,
NOT read, so CRC validity is not independent evidence
here.  Decorative graphics bands may differ between
captures and are not byte-certified.
