## The symptom
Two identical services: agent 6.0.0.154 never registers with New Relic in a
network with no default route (proxy only), while 5.24.0.153 works with
`HTTPS_PROXY`. Giving the v6 service a direct route makes it register. The
agent is ignoring the proxy.
## Why
The New Relic Python agent does not read the standard `HTTPS_PROXY` env var.
It uses its own settings: `NEW_RELIC_PROXY_HOST` (plus `NEW_RELIC_PROXY_PORT`
and related `NEW_RELIC_PROXY_*` settings).
## The fix
Set `NEW_RELIC_PROXY_HOST` to your proxy host (with the port and other
`NEW_RELIC_PROXY_*` settings as needed). Setting it resolved the reporter's
registration failure immediately.
## Related
The same issue also caused an unrelated-looking httpx instrumentation crash
(`'NoneType' object has no attribute 'process_response'`), which was fixed in
the following agent release. If you see that crash alongside proxy trouble,
upgrade the agent too.
## The general lesson
"Works without proxy, fails with proxy" on an agent that claims proxy support:
check whether the agent has its own proxy settings before assuming it honors
the standard env vars. Many agents do not.