Note that the material thickness doesn't matter; 10GHz doesn't go more than a few micrometers into most any metal. The bulk is mostly incidental (but does help when power dissipation is needed, as for PAs). Machined enclosures are used more to get the shield walls closer to the circuit. Since the circuit is composed of varying sizes of traces, filter structures, amplifiers and other components, the envelope is always varying, so, easy enough to mill everything out of a solid block and call it good.
Why the proximity? Keeps cavity modes suppressed. The cutoff frequency of a cavity is determined by the max width of its minimum cross section (or something like that?). Keeping it tight to the circuit, keeps those to higher frequencies.
Whereas the diecast box, you're probably just stuffing a board in there, with a few bulkhead connectors. The connectors might not even have great mode transitions. Which... probably, the kind like the 5-prong vertical SMAs and whatnot, are about the best? Anything open-structure or wired is NFG of course. Edge-launch, well if you can get the board into the damn thing, not too bad, but you'll be limited on board area that way, have to leave clearance to get the connector into the hole. And the board is then left free-hanging on one edge, not great mechanically or electrically.
All hope is not lost, of course; it just means you may need to invest in some RF absorbent materials for the diecast box -- particularly when placed within 1/4 to 1/2 wave of the surface (enclosure, PCB or both), this absorbs the incident plus reflected wave interacting with the wall or whatever. Note that if those fields couple noticeably with, say, resonators, filter structures, whatever on board, they'll kill the Q -- that's leakage that would otherwise be reflected in-phase with nearby shields, but here can only be absorbed, or allowed to ping around inside the enclosure, likely also coupling into other undesirable / unintended paths.
Tim