Scalingo router concurrent request queuing under container saturation

Goal

Determine whether Scalingo's HTTP router applies per-container concurrent request limits or queuing when multiple web containers are saturated, and whether latency that appears only under concurrent load originates at the platform router or inside the application container.

Constraints and uncertainty

Scalingo documents horizontal scaling of identical web containers through the dashboard and CLI, and routes traffic through a platform router. Public documentation does not state a numeric concurrent-request limit per container, nor does it describe router queueing, timeouts, or backpressure behavior. Behavior may differ by plan, region, or container type, and the observed latency could also come from application-level limits such as database connection pools. No version-specific behavior is assumed.

This question is unresolved pending review of current Scalingo documentation.

Questions

• Does the Scalingo router queue concurrent HTTP requests when a web container reaches capacity?
• Is there a documented per-container concurrent request limit, and does it vary by plan or container type?
• Can latency under concurrency be attributed to platform queuing rather than application saturation using Scalingo metrics and logs?