[Support] Nginx upstream worker_connections not being reached under high load — troubleshooting help

I have a cluster of 4 backend nodes behind an Nginx reverse proxy handling ~50k concurrent connections. Worker_processes=auto and worker_connections=1024 seems appropriate but I am seeing 'worker_connections are not enough' errors in the error log during spikes. The system has plenty of RAM (64GB) so this isn't a resource limit at the OS level — it is strictly Nginx configuration.

Questions:

• Is worker_connections per process or total? If 8 cores, that's only ~8k connections total for the whole proxy.
• Should I increase to 16384+ and tune sysctl net.core.somaxconn accordingly?
• Are there any hidden Nginx limits (worker_rlimit_nofile) that could be throttling this?

Logs: `[error]