Does automatic garbage collection impact performance in large repositories?

Managing large repositories in Tower often requires balancing repository health with operational speed. While Git provides internal garbage collection (GC) to optimize storage and remove unreachable objects, the timing and frequency of these operations can introduce significant latency during active development.

There is uncertainty regarding whether enabling automatic garbage collection leads to unpredictable performance dips during critical fetch or commit operations, or if the overhead is negligible compared to the benefits of a compacted repository.

What is the recommended approach for managing garbage collection in Tower for high-volume repositories? Does the client provide a mechanism to schedule these optimizations during off-peak hours to avoid blocking the UI?