Netting Channel State must be cleared after all unlocks are done

[hackaugusto]

Abstract

Currently the channel is cleared once the partner locksroot is unlocked:

raiden/raiden/transfer/channel.py

Lines 1792 to 1806 in 1ee26cd

	def handle_channel_batch_unlock(
	given_channel_state: NettingChannelState, state_change: ContractReceiveChannelBatchUnlock
	) -> TransitionResult[NettingChannelState]:
	events: List[Event] = list()
	new_channel_state: Optional[NettingChannelState] = given_channel_state
	# Unlock is allowed by the smart contract only on a settled channel.
	# Ignore the unlock if the channel was not closed yet.
	if get_status(given_channel_state) == CHANNEL_STATE_SETTLED:
	# Once our half of the channel is unlocked we can clean-up the channel
	if state_change.participant == given_channel_state.our_state.address:
	new_channel_state = None
	return TransitionResult(new_channel_state, events)

This logic was correct before #3553 , because the data was not needed, however after the economic batch unlock the channel data has to be preserved until both locksroots are unlocked.

Motivation

Because of that logic, when there are two unlocks to be done one of the nodes will always clear on the first unlock and the second will hit the following assert:

raiden/raiden/blockchain_events_handler.py

Lines 406 to 410 in 1ee26cd

	msg = (
	f"Can not resolve channel_id for unlock with locksroot {pex(locksroot)} and "
	f"partner {pex(partner)}."
	)
	assert canonical_identifier is not None, msg

Specification

The channel should be cleared only after all unlocks are done.

Backwards Compatibility

This will change the state representation, so a migration is needed to update the channel's state.