[Bug] RabbitMQ Failures Silently Ignored - Event Bus Breaks #4

Problem

When RabbitMQ is unavailable, services log a warning and continue without publishing events. This silently breaks the entire event-driven architecture.

Impact

• Event bus failures go unnoticed
• Workflows progress without event notifications
• Services become inconsistent
• No visibility into failures

Root Cause

File: services/matching/src/server.js (line ~32)

try {
  publisher = await EventPublisher.create(config.rabbitmqUrl);
} catch (error) {
  logger.warn({ err: error }, 'RabbitMQ unavailable; continuing without publishing');
}

Solution

• Make RabbitMQ connection mandatory in production
• Fail service startup if RabbitMQ unreachable (except development)
• Add retry logic with exponential backoff
• Implement dead letter queue for failed events

Acceptance Criteria

• Services fail to start if RabbitMQ unavailable in production
• Retry logic with exponential backoff implemented
• Dead letter queue setup documented
• Development mode allows graceful degradation

Related

Affects: All services (#1-#7)
Priority: Unblock before production