`UnboundLocalError` in `_flush_events` masks WASM trap errors

[jonathannorris]

Problem

_flush_events in event_queue_manager.py has a latent UnboundLocalError that masks underlying WASM errors.

# event_queue_manager.py:86
def _flush_events(self) -> int:
    if self._flush_lock.locked():
        return 0

    with self._flush_lock:
        try:
            payloads = self._local_bucketing.flush_event_queue()
        except Exception as e:
            logger.error(f"DevCycle: Error flushing event payloads: {str(e)}")

        event_count = 0
        if payloads:          # <-- UnboundLocalError if exception fired above
            ...
        return event_count

If flush_event_queue() raises (e.g. a WASMAbortError from the AssemblyScript bucketing engine trapping during flush), payloads is never assigned. The if payloads: check on line 97 then raises UnboundLocalError: local variable 'payloads' referenced before assignment.

The _flush_lock releases correctly (the with block handles that), but the UnboundLocalError propagates into run()'s generic except, which logs a vague warning. The original WASM error is effectively masked.

Context

Found during a cross-SDK audit related to dotnet-server-sdk #204, where the WASM bucketing engine traps during flush due to AssemblyScript throw new Error(...) (e.g. "Request Payload has not finished sending" in requestPayloadManager.ts:281).

The Python SDK does not have the same flush-mutex-leak as .NET (with self._flush_lock always releases), but it does have this related masking issue.

Suggested Fix

Initialize payloads before the try block:

payloads = []
try:
    payloads = self._local_bucketing.flush_event_queue()
except Exception as e:
    logger.error(f"DevCycle: Error flushing event payloads: {str(e)}")