Operators can set SNOWLUMA_SESSION_RELEASE_DEADLINE_MS, in milliseconds.
The wait stays at 10 seconds when that variable is absent or not a positive integer.
If the previous OneBot session does not finish closing, the account now
comes back online instead of staying disconnected until QQ is restarted.
Fixes#509
Give downstream bots a notice when an account comes up or goes down, so they can handle their own alerts. Operator notification channels are unchanged.
Friend, group, and member lists load as soon as a Bridge can send
packets, instead of waiting for OneBot database preparation. Login
history sync still waits until the instance is usable and both lists
have loaded.
Report pending group invitations through the standard OneBot request event even when their origin does not provide a live notification.
Prevent duplicate delivery when the same invitation is observed more than once, stop outstanding work with the account lifecycle, and preserve previously supported invitation behavior.
Expand group-system message queries and surface incomplete lookups instead of reporting misleading empty or not-found results.
Start each account as soon as its message storage is ready, then continue older history updates without delaying normal account availability. Keep already started accounts online when maintenance needs to be retried.
Handle preparation failures, stale completions, early exits, logout, and shutdown consistently. Add lifecycle coverage for startup ordering, automatic retries, cancellation, and account availability.
Serialize desired network plans, isolate per-adapter apply failures, restore prior bindings when possible, and report saved versus applied state. Quiesce Action and event ingress, await in-flight dispatches and transport closes, retain failed generations for retry, and propagate shutdown failures.
DrSmoothl