Pull Request Overview

This PR refactors the HTTP streaming flush mechanism to trigger immediately on the first emit rather than after a 500ms delay. The changes improve responsiveness while still maintaining rate-limiting through subsequent 500ms delays between flushes.

Key changes:

• Modified HttpStream.emit() to trigger immediate flush on first call instead of using a timer
• Updated concurrency control in _flush() using manual lock acquire/release with early returns
• Changed timer initialization from get_event_loop() to get_running_loop() for better async safety
• Added optional server parameter to HttpPlugin constructor for dependency injection

Reviewed Changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

            if i == 0:
                self._logger.debug("No activities to flush")
                return

    1. emit() adds activities to a queue and cancels any pending flush timeout
    2. emit() schedules _flush() to run after 0.5 seconds via Timeout
    3. If another emit() happens before flush executes, the timeout is cancelled and rescheduled

I think you should try replacing the sleep statements in the tests with stubbing the http client and making it throw 429 at different times, ensuring that the streamer can recover from http rate limits, then also some other error codes to make sure it behaves as expected.

Right now the sleep statements will affect the overall test suite performance but also, this doesn't accurately simulate network rate limits

Unsure why we need the polling?