docs(iter260): close unreproduced profile cleanup investigation

[ractive]

Outcome

Iteration 260's bounded investigation did not reproduce a skipped profile removal: all 20 isolated launch/stop cycles returned removed=true and skip_reason=null. Twenty separate consecutive runs of the unchanged exact cleanup test passed. The selector diagnostic test also passed with its marker/profile/PID observed during execution and no attributable surviving process afterward.

Close the plan as obsolete, unreproduced, as its original Theme A permits. Preserve the unproved mechanism and repair criteria unticked; no product/test change or speculative repair is introduced. The historical orphan explanation remains plausible, not proved.

Validation

• 20 launch/stop cycles and 20 exact gated live96 passes, distinct retained logs.
• Two exact live160 passes: first observer used the wrong profile-root nesting; the corrected repeat supplied during-run attribution. Both attempts retained.
• All nine xtask gates passed, dogfood explicitly skipped without a script; plan check, HYALO005 and diff check passed.
• Reused passing ordered source gates on unchanged relevant inputs. Before checkpoint the branch fast-forwarded to verified 261 merge38add406; its reviewed/gated source is unchanged by this docs-only PR.
• Independent local review: explicit zero findings. Supervisor finalized only status, landed-answer checkbox, evidence references and Hyalo feedback.

Existing sweep evidence

No new full sweep was needed for documentation-only work. Part B reuses the separately owed242 sweep already recorded September14: 343 exact names across five tiers, 331 passed/12 failed, zero skipped/preexisting/vanished/launch-timeout/timed-out tests and zero leaked/unattributed profiles. Both named Part B tests passed; all historical failures keep their existing dispositions. This is not claimed as a green sweep.

Carry-over

Unreproduced false-removal mechanism and hypothetical repair criteria remain unticked. Reopen on an attributable false removal with the actual skip reason and OS/profile/owner evidence. Historical245 dead-owner directories are not evidence of a stop-path cause. HN's historical readiness-versus-site cause remains unproved under its existing next-failure diagnostic trigger. No broader backlog work was executed.

Evidence: .git/ralph-loop/20260919-queue/iter260/ and independent iter260-review/; original owed242 sweep remains in the September12 run directory.