blocked (data source down after GVL reboot)

Viewer CODE is intact + live: serve.py back up on :8094 (new srv_omidlab tmux
window, .venv python), public omidlab.net/runs/ -> 200. No long-running job of
mine; nothing of mine to restart.

BUT /api/runs/list returns 0 groups. Two post-reboot causes:
1. Mount layout changed — everything now sshfs'd from 100.74.71.38 under
   self-mirrored /data/cameron/* paths. My DEFAULT_ROOTS still point at the OLD
   /home/cameronsmith/mnt/yukon/cameron/puget/{wandb,code/wandb,data/wandb} which
   no longer exists.
2. /data/cameron/yukon_remote (where the data likely now lives) is a STALE FUSE
   endpoint: "Transport endpoint is not connected". Needs remount (infra, not me).

Plan once the mount is healthy: confirm the new wandb path under yukon_remote,
repoint wandb_runs.py:DEFAULT_ROOTS, re-verify in browser.

In-flight UI request at reboot ("cap image height so wide panels pack two per
row"): held — can't visually verify or measure panel aspect ratios until the
data mount returns. Note: wide (1008px) panels are width-constrained, so a pure
height cap may not pack them 2/row; will confirm axis against real dims.