Have you ever been in this situation? It is a public holiday after a weekend, you get up and want to check if an earthquake happened during the night. You choose to check the data on the Geofon-Website (it's a good one). No noteworthy quake happened but you wonder: Why are there so many quakes happening in 10 km depth?
Is it coincidence or is there a reason why quakes prefer to nucleate in 10 km depth?
If you are using Google to find the answer, you might find some very different ones. Of course also the correct and logic answer (next paragraph) but also a lot of... nonsense. Most of the nonsense wants to tell you that earthquakes are used as a weapon that can only work in 10 km depth but there is also some even more weird stuff.
However, the explanation why (especially Geofon) lists so many quakes in 10 km depth is more simple (and boring): It's just a placeholder. Many earthquake surveys are using computer software to locate earthquakes automatically. This has the advantage that data can be processed and published within minutes, but also gives a higher probability of errors.
While the location of an epicenter and determination of the quake's magnitude is rather easy, locating the depths (hypocenters) is more difficult. Often automatically working systems are not able to provide good depth locations, especially if quakes happen in a location far away from your closest station. The Geofon-network, although rather dense, does have many places around the world with little or no station coverage, especially the ocean bottoms, where it is (even with supervision of a human expert) extremely difficult to determine the depth.
So: If you nevertheless want to publish the depth data you have two options: Use the automatic (and thus maybe highly errorness) solution or use a fixed value which is wrong but at least gives you a realistic estimation for shallow crustal earthquakes.
And here we have the Geofon-method: Giving 10 km depth automatically for all quakes, where no better solution can be determined due to the lack of data - until a human expert can supervise it and maybe give a reliable solution.
In case of the list above this method seems legit: You get a 10 km solution for quakes which are either rather small, about M4-ish (the Crete ones), in areas without Geofon station (Iran, Mariana Islands) or in areas without any station (Mid Atlantic Ridge). Plus: Today is pentecost monday, a public holiday in Germany (where Geofon is centered) and whichs follows a weekend. As long as no significant event happens, the Geofon staff does only basic and necessary work during weekends and public holidays (and nights) - this does not include hypocenter location of M4 aftershocks in Crete or quakes far away from human settlements.
Usually manual revision of this 10 km placeholder is done by Geofon on the next work day, at least for the more relevant quakes (M5+) where enough data for a reliable location are available. This is not only a Geofon-method. Also USGS and other smaller surveys are using this method, some with other placeholders like 5 km or 33 km.
If you don't get a proper depth location from Geofon (or don't want to wait), it's a good idea to consult local earthquake surveys of the affected area. These often have a higher station density around the epicenter and might thus even be able to automatically (but surely after manual revision) locate a hypocenter. In case of the list above, the Crete quakes were properly located by the University of Athens. There you see: Not only 10 km depth but 6 km, 12 km, 14 km and also a 10 km solution (which might be real in this case).
However, even these manually located depths have some km uncertainties in the range of 2 or 3 km. This shows: Depth location is not easy, not for seismologists and surely not for an automated software. But 10 km depth, as a placeholder, is at least not too far away from reality in most cases.