Writing to the Booking DB SQL Server RDS instance was failing with a generic timeout error.
After investigation, we found that INSERTS were being locked, but by sleeping SPIDs. Killing the process caused the lock to roll to a different process, in an endless rolling lock cycle.
We discovered that there was a fault in one partition of the instance, which we failed over to a new instance and rebooted, which cleared up the locks and allowed those bookings that had been made during the issue to complete their transactions and everything now looks ack to normal.