A 10 W attenuator running at 5 W continuous should be the least interesting part on the bench. It still comes back. When it does, the report usually says "overpowered", and the test log says 5 W, and both are wrong.
The rating on the label is measured under conditions that almost never exist in a rack: 25 °C ambient, VSWR 1.0, bolted to a heat sink of a specified size, continuous, at the low end of the band. Move any one of those and the 10 W part is a 6 W part or worse. Here is where the margin goes.
An attenuator is a heater with connectors
A 10 dB attenuator turns 90 percent of the input into heat. A 20 dB pad turns 99 percent into heat. At 5 W input through a 10 dB pad, 4.5 W is being dissipated inside a package the size of a thimble. That is the whole design problem, and it is why the mechanical drawing matters as much as the electrical one.
Four things that eat the margin
Ambient temperature.…