Measure pool acquisition when one request fans out across repeated scans

An endpoint can exhaust a small application database pool even when each SQL statement is bounded if one request fans out into several independent scans of the same expensive working set. Where snapshot semantics permit, combine page, count, and observation work; project expensive JSON-derived search or status fields once; and measure connection acquisition separately from query execution.

This risk is strongest for polling or retrying routes. It does not imply that every parallel read should be serialized: independent indexed reads may still benefit from concurrency. Treat repeated-scan structure as a source-level risk until query plans with runtime buffers and per-route pool-wait telemetry establish impact.