docs(home.nix): the laptop leg does not work for the reason this comment gave

[ZacxDev]

docs(home.nix): the laptop leg does not work for the reason this comment gave

MEASURED 2026-09-09, while verifying that the tmux restore stack operates on the
laptop host after #1415/#1376 shipped.

The comment on the tmux-snapshot-push unit's TMUX_TMPDIR pin said:

the LAPTOP leg gets no such guard ... It works because the laptop's socket
is in /tmp. If the laptop ever acquires this host's arrangement it silently
reports zero windows

Both halves are wrong, in opposite directions, and a reader reasoning from
either lands somewhere false.

THE PREMISE IS FALSE. The laptop already HAS this host's arrangement:

/tmp/tmux-1000                        does not exist
/run/user/1000/tmux-1000/default      is the socket
the server's own /proc/<pid>/environ  carries TMUX_TMPDIR=/run/user/1000

THE PREDICTION IS ALSO FALSE. By the old reasoning that leg should now be
reporting zero windows. It is not. Measured over the exact path session-manager
uses — a NON-INTERACTIVE ssh, which sources .zshenv only:

TMUX_TMPDIR=[/run/user/1000]
tmux list-windows -a  ->  29 windows
grep -c TMUX_TMPDIR ~/.zshenv  ->  0

So devrc still does not set it, exactly as the comment says — but the value is
in the ssh session environment anyway. That is the SAME "undeclared runtime
state" this comment already flags two paragraphs earlier for the workbench: a
fact about this boot, not a property of the configuration.

WHAT THE CORRECTION KEEPS. The failure SHAPE is untouched and still worth
knowing: if that leg cannot find the socket it reports zero windows while
session-manager records reachable:true and windows_measured:TRUE, so the
pusher's torn-collection gate cannot catch it. Only an env pin on that side
would. The hazard is MASKED, not absent — the correction says so explicitly, so
nobody reads "it works" as "it is pinned", and says not to restore the /tmp
explanation.

This is the shape claude/RULES.md names: a comment is a claim too, and this one
would have led a maintainer to the wrong conclusion in either direction — either
"the laptop is fine because /tmp" or, on seeing the socket move, "the laptop
must be broken now".

Verified: nix-instantiate --parse nix/home.nix OK; test_tmux_snapshot_push.py
37 passed; the test this comment names,
test_the_unit_gives_tmux_its_SOCKET_directory, 2 passed. Comment-only change —
no evaluated Nix expression is touched.

Co-Authored-By: Claude Opus 5 (1M context) noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_013jdbmhCKa6edhTmiADsziR