master_sync_reply in channeld assumes that there is only one such call pending at any point in time. This may not always be true, since we have 3 places in which this is called in response to timers and/or incoming messages from the peer. io_wait is used on the peer messages, so concurrent peer triggered calls should be safe, but if we have a commit timer running, followed by a call triggered by a peer message we'll run into the following error:

lightning_channeld: channeld/channel.c:518: master_sync_reply: Assertion `!peer->handle_master_reply' failed.

This'll kill channeld, requiring a reconnect.