TikTok faced significant challenges with its core payment service, originally built in Go, as user traffic surged. The service became CPU-bound, leading to performance issues and rising operational costs. To address this, the team identified the most resource-intensive API endpoints and opted for a selective rewrite in Rust. This decision aimed to enhance performance without overhauling the entire system, allowing the rest of the service to remain in Go. The process unfolded in three phases: rewriting the targeted endpoints in Rust, rigorous correctness testing, and comprehensive stress testing. The results exceeded expectations. The Rust service managed over 150,000 QPS compared to the Go service's 85,000 QPS, marking a performance improvement of nearly 2x. This efficiency not only improved latency but also allowed for a significant reduction in necessary infrastructure, projecting annual savings of about $300,000 by decreasing the demand for over 400 vCPU cores. The project underscored a key engineering principle about using the right language for specific problems. While Go remains the preferred choice for most microservices due to its developer productivity, the successful integration of Rust for high-performance scenarios highlighted the importance of tailoring solutions to the task at hand. The insights gained from this experience reinforced the value of understanding the strengths and limitations of different programming languages.