What one cycle did with the versions it examined.

Why a cycle stopped before it finished.

One cycle's result: what it did, why it stopped early if it did, and what it could not read.

Whether a cycle's counts cover what they claim to: it walked the whole store it was given, and every rule that decided read the facts it needed. Nothing about permission enters here.

Whether a halt stops the Dredger for the life of the process. Only the cap does, because a breaker that re-closes itself is not a breaker. Every other halt is re-read next cycle.

The operator-facing text of a halt, naming the backend that raised it and what to fix.

The advisory generation an audit line names, or that none was loaded.

One cycle's counts, as its closing line reports them.

A store fault as an operator reads it: the transport's own cause and its bounded detail.

Name a store without assuming whether it is the mirror or private target.

One target's standing permissions as a preview found them.

What a preview could see of one standing permission. A preview exercises none of them, so an unmet one is reported and never acted on.

Whether a real sweep of this target would pass both standing permissions.

One target's line, which a preview prints above its counts. It closes on what no read settles: a preview deletes nothing, so it proves no authority to delete.

What one cycle could not read. A count taken with a gap open describes part of the store, so it is reported apart from the counts themselves rather than folded into them.

The gap one mount deciding without an advisory generation leaves.

The gap one package the store served no metadata for leaves.

Whether a cycle read every fact its counts rest on.

What a cycle could not read, as its closing line reports it, naming only what it did miss.