2.0: Enforce event-stream ownership, cleanup, and diagnostics #36445

Parent: #36441

Problem

The generated Promise SSE client cancels its ReadableStream reader after a parser failure, but under Bun 1.3.14 that does not reliably close the underlying HTTP response. The TUI reconnect loop can open a replacement while the server retains the previous response, EventV2.liveBounded listener, and queue.

A real-socket reproduction sends malformed SSE while keeping the response open. The client rejects with MalformedResponse, but the server still reports one active response after cleanup.

Without diagnostics, active response count, event-bus subscriber count, and serialization volume are invisible until CPU and memory become pathological.

Goal

Guarantee one active event stream per client owner, prompt resource release on failure/teardown, and enough bounded diagnostics to identify leaks.

Requirements

• Define ownership for one mounted TUI and one event-stream generation.
• Abort or fully dispose the previous transport before opening its replacement.
• Ensure HTTP response finalization removes the event listener and shuts down its queue.
• Handle malformed data, oversized events, transport failures, server restarts, and consumer exceptions.
• Add retry backoff/jitter where synchronized clients would otherwise reconnect in phase.
• Expose content-free diagnostics for active streams/subscribers, opens/closes, serialized event counts/bytes, overflows, and reconnect reasons.

Acceptance criteria

• One mounted TUI has exactly one active stream and zero after teardown.
• Repeated failure/reconnect cycles never exceed one active stream per TUI.
• Eleven TUI clients produce exactly eleven active streams, not an accumulating count.
• Cancelling a real network client closes the server response and returns listener count to baseline.
• Each event ID is handled at most once per TUI generation.
• Diagnostics expose leaked/overlapping subscriptions without logging payloads.