[rayvader]

Summary

The Froyo widget no longer fails silently. A failed read now says why, and reads prefer Froyo's live list.

Read result	Header
good	N open · Updated HH:mm
login ran out (after one renewal)	N open · as of HH:mm · sign-in
no connection / 502 / 503	… · network
socket timeout / 408 / 504	… · timeout
other HTTP error	… · server
200 but no list in the page	… · format

• Rows are always kept on failure. With no saved list yet: Couldn't update · <reason>.
• Each failure is logged under tag FroyoWidget with the HTTP code or exception (adb logcat -s FroyoWidget).
• Reads prefer open_items_live.items (Froyo page v4: kept current every 15 min by its sync job) and fall back to open_items.
• A 400/404/409 drops the remembered thread and asks for it again, once.
• The dashboard read waits up to 45 s (the gateway composes the whole page per read).

Checked on the gateway side: auth on /api/members/froyo-personal-butler/dashboard passes in its audit log, and the v4 page, rendered with the gateway's own code, carries open_items and open_items_live in the data island the widget reads.

Tests

• UpcomingTest: v4-shaped response prefers the live list; falls back when it is missing, not an object, or has no items; empty live list is "nothing open"; reason table; stale-thread codes; as of HH:mm carries the reason.
• e2e Froyo widget: fake gateway serves the v4 shape with a stale extra row in open_items (never shown); 503 -> · network; 200 without a list -> · format; rows kept; next good read clears the reason.

Not covered

• The phone's own failure. The new reason and logcat line will name it.