osiris679

idk about Claude, but ChatGPT 5.5 can most definitely spot security definer issues during routine audits, unprompted

Not using the log drain feature, just logging inside each execution context, so Edge Functions, cron jobs, and RPC paths all emit structured start/success/error logs with shared metadata like request/run IDs and stack traces, then centralize those events into one logging pipeline/table and propagate a correlation ID across edge -> cron -> RPC to trace a full flow end-to-end; that’s why we can see failures as they happen instead of only after invocation completion

This only works if you don't rely on socket streaming for "live" DB tables, or other difficult operations to proxy with, otherwise this solution becomes more unreliable at scale.

Would using the custom DNS domain name resolve the issue for end users? Or is there a block on deeper packet inspection?

On which part? Happy to detail 

No, this view: [https://supabase.com/dashboard/project/{project-id}/logs](https://supabase.com/dashboard/project/{project-id}/logs) with the new beta features enabled

I use Sentry to log caught and uncaught errors in my edge functions, CRON jobs, RPC function calls, etc., along with Discord channel notifications. This is another layer on top of that observability.

I'm not sure what your product is and the technical requirements involved, but have you calculated the cost/benefit of Supabase alternatives as you scale? For instance, who will own backups, monitoring, migrations and scaling once the database is on Railway, and how much extra complexity will that add for a growing team compared to the abstraction Supabase currently gives you? I don't think it's a bad idea to outgrow from Supabase, but I want to make sure you've considered the long term as you build your product.