Two bugs on the propagation map:
1. LiveStash reconnect path crashed the LiveView with KeyError on
:initial_scores_json. stash_assigns/2 only persists selected_band and
selected_time, but the recovered-socket branch returned early without
rebuilding any of the other mount-time assigns (bands, bounds, the
initial score JSON payload, grid/radar toggles, antenna height).
Users on the Phoenix longpoll fallback — or anyone whose websocket
reconnected after a brief network blip — saw a blank page and a
server-side 500. Refactor mount/3 to always compute ephemeral assigns,
using recovered selected_band / selected_time when present and
falling back to defaults otherwise.
2. The propagation reach polygon shown after clicking a point never
redrew when the user changed bands or scrubbed the forecast timeline.
The redraw logic only ran inside the point_detail handler and was
gated on `hull.length >= 3`, so the stale polygon from the old band
stuck around whenever the new band had no coverage at the clicked
point. Extract a redrawReachPolygon/0 helper on the Leaflet hook that
always clears the previous polygon first and pulls the tier color
from the current score at the clicked location, then call it from
update_scores (band change + server time scrub) and the cached
timeline-scrub path. The point_detail handler delegates to the same
helper so the polygon stays in sync with both the score grid and the
detail panel.
Graham McIntire
eedd91c6ee