SECURITY FINDING 1 on the review-fix-1 commit (875c339c) — the hub found its „first sight of a large request” by searching window text that also held box-supplied names.

### RED: the new tests before the fix (felhom.eu be272b7c)
=== RUN   TestD84_ABoxCannotPlantTheLargeRequestMarker
    ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "more than half of the store (apps: ghost)"
    ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "more than half of the store"
    ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "the request is for 20 of 20 snapshots — more than half of the store (apps: ghost): it waits one week, and the operator is told"
    ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "large:ghost"
    ended_large_marker_test.go:45: the box planted the marker a week ahead and the hub agreed to 10 ids at once — no wait, no warning
--- FAIL: TestD84_ABoxCannotPlantTheLargeRequestMarker (0.05s)
=== RUN   TestD84_TheLargeRequestRecordIsBoundedAndGoesWithTheHousehold
    ended_large_marker_test.go:70: 500 different requests left 0 rows (limit 50): <nil>
--- FAIL: TestD84_TheLargeRequestRecordIsBoundedAndGoesWithTheHousehold (0.04s)
FAIL
FAIL	gitea.dooplex.hu/admin/felhom-hub/internal/offsitekeys	0.103s
FAIL

### After the fix: mutations
== MUTATION: a claimed name that is not an app name is written into the ledger
   RED on the mutant, GREEN on the real file
     --- FAIL: TestD84_ABoxCannotPlantTheLargeRequestMarker (0.05s)
     ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "more than half of the store (apps: ghost)"
     ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "more than half of the store"
     ended_large_marker_test.go:37: a claimed name that is not an app name was written into the window's row: "the request is for 20 of 20 snapshots — more than half of the store (apps: ghost): it waits one week, and the operator is to
== MUTATION: the hub does not record its own first sight
   RED on the mutant, GREEN on the real file
     --- FAIL: TestD84_ALargeRequestWaitsAWeekAndOneWindowAgreesAtMostTheCapsSize (0.09s)
     ended_bound_test.go:108: a week later: 0 ended ids agreed with a cap of 10 — want exactly the cap's size
     --- FAIL: TestD84_TheRestGoesInTheNextWindow (0.04s)
     ended_bound_test.go:161: a week later: 0 agreed, want the cap's size 10

2 mutations, 0 not a red-proof
== MUTATION: the first sight is read across households
   RED on the mutant, GREEN on the real file
     --- FAIL: TestD84_ABoxCannotPlantTheLargeRequestMarker (0.07s)
     ended_large_marker_test.go:59: another household reads this household's first sight
== MUTATION: a first sight older than the memory still ends the wait
   RED on the mutant, GREEN on the real file
     --- FAIL: TestD84_ABoxCannotPlantTheLargeRequestMarker (0.05s)
     ended_large_marker_test.go:56: a first sight older than `since` was returned
== MUTATION: no limit on one household's rows
   RED on the mutant, GREEN on the real file
     --- FAIL: TestD84_TheLargeRequestRecordIsBoundedAndGoesWithTheHousehold (0.60s)
     ended_large_marker_test.go:74: 500 different requests left 500 rows (limit 50): <nil>

3 mutations, 0 not a red-proof
== MUTATION: the record does not go with the household
   RED on the mutant, GREEN on the real file
     --- FAIL: TestD84_TheLargeRequestRecordIsBoundedAndGoesWithTheHousehold (0.26s)
     ended_large_marker_test.go:74: 500 different requests left 0 rows (limit 50): <nil>

1 mutations, 0 not a red-proof
