This will break for > v5!

This was intentional, just in case Fastify v6 uses a different diagnostics channel hook for errors when published. It won't break if 6.x is used here, in the worst scenario, the error will be captured twice, and deduped eventually.

I'd rather we check if our tracing:fastify.request.handler:error diagnostics channel listener was registered, that feels like a more robust check.

I guess this is hard because others can also register on tracing:fastify.request.handler:error. Perhaps we can do this by setting a field on the integration? Then you can grab the current client fastify integration instance and check.

I have just tried that. Looks like it's a bit tricky because dc.subscribe does not return anything (we don't know if it subscribed or not at instrumentation time. If the hook is not found, it fails silently). We can set a flag in dc hook when an error happens, but the error handler plugin will be registered already, then we can maybe unregister it. Wdyt?

We can set a flag in dc hook when an error happens, but the error handler plugin will be registered already, then we can maybe unregister it

yeah lets give this a try