Shipwreck hunting is ridiculously expensive. The resources required to exhaustively explore 100 sqm of space is probably 1000x of the resources required to do it on land. There aren't any easy shortcuts: radar doesn't work underwater, sonar does but is extremely low resolution, lidar works pretty well but only if the water is very shallow and clear, underwater drones have extremely limited mobility and communication capability. A lot of funding in archeology tends to go to easier or higher probability wins, which has mostly been aerial lidar in heavy vegetation areas for the past 10-15 years.

The best shipwreck hunters rely almost entirely on probabilistic models for where they might find shipwrecks, and the most useful probabilistic models have all developed in the last 30-40 years. In fact, some of the best probabilistic models like Bayesian Search Theory actually originated as a formalization of heuristics that were already used in treasure/shipwreck hunting.

In that respect, I would argue that this find is actually the result of recent advances in probabilistic modeling (along with other advances in data engineering with respect to extremely messy historical data sources) that have just barely gotten accurate enough to start getting the funding it needs to do the harder work of actually working on the sea floor.