One problem with these hypersonic endeavors and DOD development in general is there are too few tests (And done too slowly) to iterate. The developers do not get lots of opportunities to learn from real-data, like at SpaceX. Instead, there are low-single-digit numbers of tests with tons of political attention ready to pounce at the smallest (And most common!) failure.
But at least with civilian rockets all the competitors have clear success metrics: they all launch with some regularity, and it's obvious if those launches were a success. Lots of weapon systems have one or two tests and then take a decade before they see any action, if at all.
Fore example the Patriot system is operational since 1981, but when first used in combat a decade later its accuracy came under heavy scrutiny, and continues to look suspect on a per-missile basis [1]. But that's one of the weapons systems that actually sees active combat, and sees improvements based on that. Now think of all the systems that are deployed but never used.
1: https://en.wikipedia.org/wiki/MIM-104_Patriot#Operational_hi...