OPERATOR DIRECTION, verbatim: "To make sure we don't muddle the in process
deployment, queue all the tailscale decisions to the end of the step. When you
bring them back up give me decisions choices with your recommendation for
each." All four sub-decisions in gap register item 21 are deferred; when
re-raised they come back as choices, each with a recommendation.

MEASURED, at the exact scope dc-mirror.sh:177-181 uses (main,restricted,
universe,multiverse / amd64 / nosource), summing Size: across each component's
binary-amd64 Packages.gz:
  main        859,001,998   338 pkgs
  universe    211,685,404   123 pkgs
  restricted  0             (empty in backports)
  multiverse  0             (empty in backports)
  TOTAL       1,070,687,402 = ~1.00 GiB, 461 packages

Context: the existing dc0 mirror is 951 GB with 1.8 T free, so backports is
~0.1% of it -- capacity is a non-issue.

Sanity-checked rather than trusted: the five largest entries were listed to
confirm the Size: field was read correctly, and a separate binary-all index
probes 404, so arch:all is already inside binary-amd64 (301 of main's 338) and
is neither double-counted nor missed.

Worth knowing before choosing: ~462 MB of main's 859 MB is LibreOffice -- about
46% of the suite is a desktop package set these nodes will never install.

Transfer estimate stated as the weak measurement it is: one single-stream fetch
measured 932 KB/s, putting 1.07 GB at roughly 19 minutes. Treat as an upper
bound; debmirror parallelises.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HvCyrwvYTTcDYnRErfMsNf