Why a validator node should always be an archive node? What are the risks and downsides of not doing so? I understand that it is probably interesting to have at least one archive node, but why all of them? If the blockchain data is kept (extrinsics and block headers), any state at any point of time can be reproduced if needed, right? So why to keep in DB the state for each block?
This question popped up when I tried to run a validator in pruning mode --pruning 256
and I got this error:
Error: Input("Validators should run with state pruning disabled (i.e. archive). You can ignore this check with '--unsafe-pruning.'"