Reliability

[fernandomg]

The main goal for this is to focus the work towards reliability and let the speed as a secondary milestone.

Demo (mini) post-mortem

During the demo at the L2 Working Group - Call #3, the intent was not able to go through.

So here's the detailed step-by-step of what we understood what happend:

1. the listener subscribes to events using ethers;
2. the subscription is via polling;
3. this listener uses the first RPC that is being passed via the provider;
4. that RPC failed (as it does quite commonly with arbitrum RPCs) during the poll;
5. Open event missed.

Posible solutions

1. Poll for events manually, define a range of blocks to gather the info form, and move the pointer upon completion.
2. Change the current architecture to allow multiple listeners to be running in parallel (multiple RPCs subscriptions, Subgraphs, ponder, lite nodes, etc); use a message broker to enqueue the mssages and let the solver be the worker.
3. Some other...?