As compact-ish explanation: A "standard" wideband RF system in an EW or RF reconnaissance platform covers between 0-18 GHz (DC up to the Ku band), or at least as much of it as possible (and Ka/mmW becoming common on new systems); and they have challenging requirements compared to a communication system. Communication systems are simpler to design since both sides of the link cooperate and filter out a wide swath of potentially interfering signals, but a military system wants to see as many signals as possible so they can be collected or jammed. It has not been advantageous to use an integrated RFSoC in the past given the requirements. If a company were spending millions of dollars designing a complicated front end, they might as well pick a separate ADC/DAC that maximizes the performance they cared about, rather than go with the "easy" integrated RFSoC option that might not have the absolute best performance. Now the industry is just getting to the point that a system like a direct sampling ADC/DAC integrated into a Versal might be able to process massive bandwidths at high enough bit rates that they can do useful things for military applications; it may actually be worth it now because you can push to really high data rates, and the additional processing might make up for a small loss in ADC/DAC performance. Give it a couple years for these to make it into new designs and get fielded.
So I guess the tl;dr, is that it is not because defense doesn't like integrated packages, they just haven't been worth it considering the design goals. Defense does move slow, but this is more about being able to field "military-grade" solutions that work well in challenging RF environments, and once that is possible the government will start to pay for it.
