tfa: "This, in turn, provides very high fidelity 'tracks' of targets thanks to telemetry from various sensors operating at different bands and looking at the same target from different aspects and at different ranges."
My take from last article: Let me tell you: fast tracks moving at insane speeds with sometimes-faint signatures are the norm. a perfectly calm sea, esp near shore, can create 100s or a 1000 spurious tracks if automated tools are trusted. The incorrect association of signatures to hypothesized contacts can produce "jumps" in their location, altitude that makes it look like inhuman maneuverability.
Second, the proliferation of small aircraft does bias operators (and algorithms) to assume a real contact from small signatures, esp at low altitudes where noise is greatest.
In a well integrated, multi-vehilce sensor network, a small false positive rate for each individual recognition and classication module produces enough bullshit to trick automated trackers into labelling targets. If we cant assume common sense filtering like "aircraft cant go from hover to mach 3 in one sweep", then the possible associations of sensor data to hypotheses is astronomical, and bullshit propegates.
Its a mess when you integrate so many sensors and systems and try to make sense of the resulting noise.