GitHub Codespaces proxy limits versus application connection pool sizing

Effective concurrent connection ceiling

The GitHub Codespaces proxy layer is widely reported to enforce a hard ceiling of approximately 100 simultaneous TCP connections per forwarded port, though the exact figure is not publicly documented and may vary by account tier. This limit supersedes any application‑level connection pool configuration; exceeding it results in dropped or rejected requests regardless of pool size.

Idle timeout interaction

Codespaces containers are observed to terminate after 30 minutes of user inactivity, forcibly closing all connections before the application's own idle timeout would take effect. This interval can vary by organization policy and container configuration, meaning long‑-lived pools may be reset independently of application‑defined keep‑alive settings.

Post-create script sequencing and proxy throughput

If a devcontainer.json post-create script launches multiple services simultaneously, the proxy may drop packets once the concurrent connection count approaches its threshold, causing brief startup failures. The proxy's throughput limit is per‑port and not configurable via devcontainer.json; only changing the container's network mode or port selection can affect it. Adjusting pool size alone will not mitigate this, and since devcontainer.json resource or pool‑size changes require a full container rebuild to take effect, iterative tuning through post-create scripts cannot instantly alter the effective ceiling.

Verification and recommended steps

• Check container logs for "proxy error: connection limit exceeded" messages after starting the application.
• Run ss -tuln or netstat -anp inside the container to count established connections on the forwarded port during peak load.
• If the count approaches the suspected ceiling, reduce the application pool size or throttle service startup to stay below that threshold.

Missing diagnostic detail: What is the observed concurrent connection count on the forwarded port when your application starts up? This determines whether the proxy ceiling is the bottleneck or if application‑level tuning is needed.