Latency spike in Detaspace Query Engine

Goal

To understand the cause of a measurable latency increase observed only when multiple clients issue queries concurrently to the Detaspace Query Engine.

Constraints & Uncertainty

The spike has not been documented in any official Detaspace release notes. Early observations suggest it may be linked to internal caching or lock contention, but no explicit error string or exception is emitted. The design choice between a per‑thread cache and a shared cache with explicit invalidation remains unresolved, and it is unclear whether the current behavior is an intentional throttling mechanism or a regression.

Questions

• Is the latency spike caused by cache contention or by an explicit throttling feature in the Query Engine?
• What locking strategy does the Query Engine employ when multiple threads update the shared cache?
• Will future releases address this concurrency‑related latency, and if so, under what version?