From 3c44cd0c67b5a88aa7b2846d1a5304f18334db26 Mon Sep 17 00:00:00 2001
From: Teknium <127238744+teknium1@users.noreply.github.com>
Date: Sun, 23 Aug 2026 00:25:35 -0700
Subject: [PATCH] docs: record why the reap grace window exists in the reaper
 docstring

Follow-up for salvaged PR #91994 -- without this note a future cleanup
pass could read the age gate as dead weight.
---
 hermes_cli/dashboard_procs.py | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/hermes_cli/dashboard_procs.py b/hermes_cli/dashboard_procs.py
index c07e9232a3..74a950db8b 100644
--- a/hermes_cli/dashboard_procs.py
+++ b/hermes_cli/dashboard_procs.py
@@ -882,6 +882,13 @@ def _reap_orphaned_desktop_local_serves(
       by another client/machine* which legitimately sit at ppid 1 after sshd
       exits. Killing those is a production incident, not cleanup.
     - never fixed-port remote serves (e.g. ``--port 9119``)
+    - never a candidate younger than ``_REAP_MIN_AGE_SECONDS`` (or whose age
+      cannot be determined). The Desktop client writes ``backend.lock.json``
+      only after the backend reports HERMES_BACKEND_READY, so during
+      concurrent multi-profile startup a live sibling is briefly unowned and
+      otherwise indistinguishable from a corpse; sparing young processes
+      closes that mutual-reap window. A genuine corpse merely waits for a
+      later scan.
     - best-effort; failures never raise to the caller
     """
     import signal as _signal