Despite these uncertainties, the instruments did indeed hear our distant acoustic calls. The first several recoveries went well, and the data look excellent. However, we then encountered a string of three consecutive instruments that refused to budge from the seafloor. All responded to our pings, but they could not lift off the bottom. Stuck in the mud? Flooded? After spending several hours attempting to recover them, we sadly moved on. Forensic analysis of some of the recovered instruments revealed a likely cause: bad AA battery cells in the release systems, such that they have power to communicate, but not quite enough to complete the release process. Recognizing this, we devised a power-saving routine for the remainder of the recoveries; sending release commands in brief bursts that used minimal power, and then waiting in near-silence until the OBS appeared at the surface. This proved successful, and in the end, we recovered 30 of 34 OBS, and 2 of 3 MT instruments. We are frustrated by the losses, but thankful for the data in hand. Next year we will return to recover the broadband OBS that are recording the earthquake data – in the meantime, we hope to engineer a new plan to coax the four missing OBS back to the surface.