Adrian Sutton

Agreed, Teku will also keep the `--subscribe-all-subnets` but it will require more and more resources to keep up with that as the number of validators grows. Also given that validators...

> How so? There's a cap on the number of attestations / epoch in the system, and we've reached peak-attestation on mainnet already. Every validator attests every epoch so if...

Looks like you need to run spotless though. :)

This error is saying that the Teku beacon node couldn't send an event to some client connected to the `/eth/v1/events` endpoint. I'm not entirely sure I understand the situation but...

ah, yeah ethereum-metrics-exporter hits the event stream pretty hard. There's a good chance it wasn't able to read the data being sent fast enough. Teku should handle this gracefully and...

The key reason you'd use a fallback in this case btw, is to handle the case where the EL is syncing or offline. Obviously if the primary beacon node is...

I'd probably go with the VC change as we can't assume the beacon node will persist the data to disk either (the api doesn't require it and I think even...

This is a good idea. It _could_ be done in the beacon node, but this kind of feature would ideally live in the MEV sidecar (eg mev-boost) which is already...

hmm, that definitely is unexpected. It doesn't happen with a simpler command line (eg `teku --network goerli --ee-endpoint http://localhost:8551 --validators-keystore-locking-enabled=false` is totally fine), but I can reproduce it with that...

Simplest version I can find that reproduces this is: ``` teku \ --ee-endpoint=http://execution:8551 \ --validators-keystore-locking-enabled=true \ --validators-keystore-locking-enabled=false \ --builder-endpoint=http://mev-boost:18550 ```