Investigate consensus steps timeouts about celestia-core

[evan-forbes]

sure thing! since starport hasn't upgraded to v0.46 yet, and we want to move away from their stuff anyway, I've been replacing the code we rely on from them with local infra, and a huge bonus to that is that we can easily change the default config.

[evan-forbes]

We had to increase the timeout propose significantly, to 10s, to come to consensus on the first round. This has other consequences, such as inconsistent block times that we will need a dynamic timeout commit to fix. We should be able to decrease the timeout propose after we increase the efficiency of tendermint's p2p stack.

I'll continue to leave this issue open, at least until we have stable defaults picked.

[liamsi]

@evan-forbes can we hardcode/update the default config such that every node locally will have the same timeouts to establish a 30s block time? This is only a bandaid until we can either upgrade to a tendermint version that makes this a consensus param, or, we backport this before launch (should open a separate issue for the latter).

[evan-forbes]

While we seem to have a working version for this by increasing the timeout-commit, I still think we should leave this open and do more experimentation on optimal values. I think doing this after incentivized testnet makes the most sense, since we have a working version as is.

[evan-forbes]

I think this issue can be closed as completed since the timeouts on mainnet seems stable. However, that is definitely not to say that the timeouts should not be changed in the future.