"The seamount last erupted in April 2011. Scientists discovered the eruption by accident, on a routine expedition to the seamount in late July."

What is the point of having all those instruments on it if they have to discover eruptions by accident. This does not seem consistent to me.

The 2011 eruption they discovered by accident.

They'd monitored the volcano before. They'd left equipment there in 2010 and in 2011 they were returning for the purpose of collecting it. The equipment monitored the volcano. They weren't monitoring the equipment.

A lot of monitoring equipment doesn't report real-time data, esp. in remote locations when you're trying to get a lot of little data points or study things over time. You simply don't need real-time data, which can be very expensive (even if in a multiply-connected world the price is slowly coming down). You plop the equipment down and then return to collect your equipment and data later. I knew a geophysicist grad student who spent his summers doing this: he'd start in June and make a circuit around recording stations. Scores of them. He'd maintain and repair them, install a few new ones every year. None were t connected to a master recorder anywhere because then they'd be constantly troubleshooting bad connections. The in-place equipment was time-tested and reliable and recorded the data for collection. In some locations he'd spend 20 minutes, swapping out a hard-drive and running diagnostics. In others, he'd spend a few days. By mid August he'd have finished his tour from Oregon to N. Calif. over to Wyoming and back through western Montana to Washington state and home to Oregon, with hard drives full of data waiting for download.

They'd sort through the data in September through December that had been collected over the previous year. Quick results during the early spring conference season for graduates just hitting the job market, first papers and posters. More detailed results came in over the next year or two, merging new data with old. They were building quite a picture of mantle structure and dynamics over the course of many years for that part of the N. American plate. In some cases, a year's data barely produced a fuzzy, clouded view of a structure. Add in years' more data and the outlines came into better focus.

Notice that the Axial Seamount 2011 was detected in 7/11, during the summer when geophysicists have 3 months off to do serious data gathering work.

When you have an ongoing, active site, that's when you invest money in the tech to provide real-time data. And when it's either economically and politically important, or scientifically red hot, that's when you get the money for real-time monitoring.

IIRC, there is an active real-time monitoring network run by the UO on the Juan de Fuca plate, but that's part monitoring the Cascadia subduction zone for stress buildup to try to predict the next megathrust quake in the Pac NW. The subduction zone is fairly close to the coast, while the spreading center with this particular seamount is hundreds of km off the coast. The last such quake was around 1700 AD. The tsunami was large enough to devastate parts of the Japanese east coast. And it hit without warning, since the magnitude 9 quake (or thereabouts) was too far away to be felt in Japan. (We stress out over 4.0 quakes. Magnitude 9 is 100 000 bigger.)