AcquireFailurePointTooOld when --start-chain-from point is past k

[ch1bo]

trafficstars

We might have a bigger problem on our strategy with --start-chain-from: Today, we wanted to start a hydra node from the beginning of the devnet to recover/see an already open head (after having an exception, other problem) using --start-chain-from and we got an AcquireFailurePointTooOld failure from the node. Searching from this in the network/consensus codebase yields: https://github.com/input-output-hk/ouroboros-network/blob/a65c29b6a85e90d430c7f58d[…]/src/Ouroboros/Consensus/MiniProtocol/LocalStateQuery/Server.hs

-> Does this mean we cannot use the acquire to "look" past k blocks? I think we haven't tried that ever on the testnet.

This should only affect the LocalStateQuery protocol, but not the ChainSync protocol.

Scope of fixing this bug could be to investigate & eliminate the need for local state query when starting from a chain point and/or use the tip for parameters etc. while synchronizing the old chain?