Two missing pages, and a schema that assumed one image per act
Steve pulled the two frames the earlier entry asked for. Both were the next in sequence.
St-Philippe, 1787 (the high-resolution register): the act ends « …de Joseph et Urbain Compte, frères de l’époux; de Charle Lerigé père, Charle Lerigé frère de l’épouse, et de plusieurs autres dont les uns ont signé et les autres, ainsi que l’époux, ne sçavent signer ». Three things settle: Pascal had brothers Joseph and Urbain; his father-in-law Charles Lérigé was alive and present, so the doubtful word before his name is du Sieur, not de feu; and Pascal could not sign — his grandfather had signed in 1751. Two more Comtes have names and no files. They stay that way until a question needs them.
Ste-Anne, 1721: the witnesses — Vinet, Pilon, Ranger, all Bout-de-l’Île families — and nothing else. The forename of the groom’s father and the day of the month are still unread, and Steve confirms that his download method gives no larger frame. So they stay unread. What fixes them is not a better guess but a different document: the parish’s other register, or Pierre’s baptism.
The tool met its first limit. A source record holds one image:. An act across two frames
is one document with two. We filed the second frame beside the first as S-0003b_… and
S-0005b_…, hashed it in the record body, and wrote the pain down in TODO.md. The rule we set
for ourselves was: a pain gets a tool the second time it appears. This is the first.