event service does not send ready/handshake immediately for newly registered or reset dispatchers when no new resolved-ts arrives, causing dispatcher initialization or recovery to stall

[zier-one]

What did you do?

Registered a new dispatcher, or reset an existing dispatcher after it had reached the ready stage, while keeping the upstream idle so that no new events advanced the resolved-ts after the register/reset.

What did you expect to see?

The event service should send a ReadyEvent immediately after a dispatcher is registered successfully, and send a HandshakeEvent immediately after a reset switches the dispatcher to a new epoch, so that the event collector can complete the register/reset control-plane transition without waiting for new upstream traffic.

What did you see instead?

Both ReadyEvent and HandshakeEvent were coupled to the onNotify() -> scanReady() path. When no new resolved-ts arrived:

• the collector never received ReadyEvent during initial registration, so it could not issue the follow-up reset request;
• after reset, the new epoch never received HandshakeEvent, so the collector could not initialize the new event stream.

As a result, the dispatcher could remain stuck in an uninitialized or post-reset unhandshaked state until another upstream resolved-ts happened to arrive.

Versions of the cluster

master