STOP 1 — operator ruling
========================
Recorded: 2026-08-18 ~09:25 UTC (11:25 CEST)
Given by: Viktor (operator), in session.

CC's recommendation was: DO NOT UPGRADE for the stated reason.
Basis: the full changelog range 4.2.2-1 -> 4.2.5-1 (128 lines, all three
entries read) contains NOTHING touching connection handling, descriptor
lifetime, accept(), CLOSE-WAIT, keep-alive or the proxy daemon's socket
lifecycle. An explicit keyword sweep returned exactly one hit, and it is a
false positive ("S3 ... honor the node's proxy settings" = HTTP proxy config
for S3 requests, not the PBS proxy daemon).

OPERATOR RULING: PROCEED WITH THE UPGRADE ANYWAY.
Stated reason: rehearsal value -- "see how that works for us, we need
practice with that too". The upgrade is therefore being performed as a
supervised practice run of the upgrade procedure on a Tier-2 protected
machine, NOT as a fix for R-336's leak.

CONSEQUENCE TO CARRY INTO THE REPORT, so it is not later misread:
this upgrade is NOT expected to change the fd slope. If the post-upgrade
slope differs, that is a surprise requiring explanation, not a confirmation
of anything -- the changelog gives no mechanism by which it should improve.
Equally, if the slope is unchanged, that is the EXPECTED result and is not
evidence the upgrade failed.