Near-field probing helped us pass the chamber twice, but only after we stopped treating the scan as a prediction of radiated emissions and started using it as a fault-location tool. The single most useful change was to scan at a fixed height with a small H-field loop, then repeat the same scan on a known-good board. A hotspot that is 10 dB above the reference board matters; an absolute number on its own usually does not, because probe coupling changes with height, loop size, and the exact placement over the trace.
I also learned to separate magnetic and electric sources before fixing anything. A small loop finds current loops and return-path problems, such as a split ground under a switching regulator or a connector shield that is only grounded at one end. A small E-field probe is better for slots, apertures, and floating metal. When a hotspot follows the loop probe but disappears with the E-field probe, I look at the return current first; when the opposite happens, I look at the shielding and the aperture. Fixing the wrong one wastes a board spin.
Two practical tips: first, mark the hotspot with a photo and a frequency, then re-scan after each single change, because two simultaneous fixes make the result impossible to interpret. Second, check the cables early. In our case, a hotspot on the board turned out to be common-mode current on the power cable, and a ferrite and a better cable ground fixed the chamber result while the board scan barely changed. Which frequency range is failing for you, and is the hotspot on the board itself or does it move when you change the cable routing?