`discoverIssues` dropped an issue this space had already taken up before anything was fetched — a saving that made a mostly-imported repository cost one appview call and no PDS reads. The rework that turned the section into goal-shaped rows made that saving wrong: the fold holds an imported issue's URI and nothing else, no title, no body, no filer, so a look that never fetched one could not draw its row at all. The imported badge, the links to the goals it became and "Import again" were therefore reachable only in the tab that did the importing, in the seconds after it, and only until the page was left. To anybody arriving afterwards the issue had simply vanished, which on an issue list reads as somebody closing it on the forge. Every open issue a look can read now comes back as a candidate. `taken` no longer decides *whether* an issue is read, only the order the bounded budget is spent in — the ones nobody has taken up first, since that is what a backlog is for — so an imported issue is dropped only when there is no room left. It is read from its filer's own repo like every other row, the index still deciding nothing. `Discovery.imported` goes with it: the count beside the section head is now taken off the rows, so an issue taken up a second ago and one taken up last week are one fact. The empty list loses its "every open issue is already a goal here" branch, which an imported row can no longer reach. Tests: `issues.test.ts` asserts an already-taken issue is fetched and offered, and that the ceiling spends itself on what is still waiting; the page test mounts with the goal already in the fold and asserts the row is drawn, badged, linked and offering "Import again" — the cold path the previous test could not see, because it only ever exercised a goal arriving after the look. Co-Authored-By: claudebot.disnetdev.com (did:plc:n6ku5xddiuguwze3f356evla) <claudebot.disnetdev.com@noreply.radial>