A few years back my layout package omitted a thermal relief on the pad of an RF part placed with a ground pour. The software is supposed to put small traces from the pad to a pour of the same net wherever it can. For some reason, when I did a minor rev of this board, the software omitted one of the thermal reliefs. This resulted in a pour ground connection on one of the two ground pads on an RF connector.
At DC this is fine. At 2.4GHz the effect is negligible. Between 5GHz to 6GHz, the error results in 8 to 10dB of signal loss. If you simply scrape away a bit of the soldermask and bridge the pad to the pour, the footprint works perfectly. It is amazing that that tiny mod results in 10 times the power being delivered to the antenna.
I suspect the reason for it has such an impact is the center conductor goes to a coplanar waveguide, i.e. a trace with grounds below it and on both sides of it. The ground plane on the left side is connected to the ground on the right side through ground vias, but there is no path not through a via. If there had been copper on the same layer as the connector connecting the two sides together, the missing thermal relief would not have mattered as much.
The cautionary part of this story is the length of the coplanar waveguide is less than 1 cm. The fact that the length of the transmission line is shorter than the wavelength being transmitted does not make it immune to transmission line problems. The entire path through the connector and connected coax is longer than a wavelength, and a discontinuity at one point can drastically effect power transfer.
This all happened years ago, but every once in a while someone new at the client asks me why we rev’ed this board, and I have to go over this infamous mistake.