Several real-time applications rely on dynamic graphs to model and store data arriving from multiple streams. Providing both high ingestion rate and efficient analytics with transactional guarantees is challenging, even more so when updates may be received outof-order at the database. In this work, we propose HAL, a novel in-memory dynamic graph database design, addressing these challenges. HAL outperforms comparable systems by a factor of up to 73× in terms of update processing throughput and up to 357× for analytics, while being the first to support out-of-order updates.