Hydra node resets head state when replaying close of another head

[ch1bo]

Context & versions

Any recent version, likely 5642332 and master

Steps to reproduce

1. Initialize a head on a network and initialize in block x+1
2. Open & close that head again
3. Start the hydra-node with --start-chain-from on block x
4. Open another, new head and do some transactions
5. Restart the hydra-node with same (start chain from x) configuration
6. The hydra-node has thrown away it's layer two state

Actual behavior

The hydra-node replays the L1 history. While most chain tx observations are ignored, the CloseTx from the old head will have the node evolve it's headState to first follow the old head getting closed and then the new head beeing initialized. Ultimately it will appear as the hydra-node has thrown away it's L2 state and got reset with a "fresh" open head.

Expected behavior

Observes and deliberately ignores Hydra transactions from other heads. For example using the headId