Actually such systems can be built, I saw a demo of one based on UWB tranceivers, it was proposed replacement for the JSTARS system. The reason it worked was that 'dumb' jammers needed to raise the entire noise floor, and 'smart' radios were using singletons (single pulses of RF energy across a wide spectrum of possible frequencies) so a small transmitter could use kilowatt nanosecond pulses which always landed above the noise floor.
If I get you right this beat on the 'dumb' jammer because the power imbalance worked to the advantage of the defender: knowing when to listen for the very short pulse at high power level coming from the transmitter and ignoring its input the rest of the time other than to establish a baseline noise level, whereas the jammer would have to blanket the whole spectrum not knowing when the next pulse would be until it came?
So essentially the only way to break the system would then be to figure out ahead of time when a pulse would be transmitted and presumably the sequence of intervals between pulses was sufficiently hard to predict that this would take much longer than it took to get to the point of the next transmitted burst?
That was pretty much the way I understood it to work, I didn't get to see 'behind the curtain' sadly. Two bits of the 'secret sauce' were keying sequencing / syncing and basically a frequency agile transceiver that operated at really high frequencies across a very wide chunk of spectrum.
The claim was that the only "known" way to jam the system was to raise the noise floor to the point where the pulses could not be distinguished, and they showed a number of scenarios that didn't work.
