The list rows only knew the manual-check states, so during an outage the
selected proxy just said 'connecting' with no hint WHY (the avito log:
19 silent handshake drops looked identical to a slow connect). Each
MTProto row now surfaces the EndpointHealth snapshot: an active cooldown
with a countdown and a human reason (handshake blocked / does not relay /
no response / DNS), a recovering state after the cooldown, and a green
'tunnel active / was OK' once main-use relay is proven. Health is polled
once a second while the box is open and never overrides an in-progress
check or the online state. Proxy choice stays fully manual.
29a2f2b16b