(The speculative SpaceX suborbital rocket thing, IMHO, is not credible evidence against my point. :) )
It is less a tech problem than a political one.
A stealth ICBM on the other hand... now that is a scary. If any other nations got wind of that we would have another nuclear arms race.
Stealth may help avoid interception during the midcourse phase, though:
https://media.nti.org/pdfs/10_5.pdf:
> Midcourse Phase – “Largest Intercept Window”
> The midcourse phase allows the largest opportunity to intercept an incoming missile. At this point the missile is no longer under power, so it follows a more predictable path. Depending on the interceptor launch location, multiple interceptors could be launched, with a delay between them to see if the first ones were successful. Since the interceptor has a longer time to engage, fewer interceptor sites are needed to defend larger areas.
> Unfortunately, a longer period in space provides an attacking missile the opportunity to deploy countermeasures against a defensive system. However the defensive system also has more time to observe and discriminate countermeasures from the warhead.
But I think you're right that it a stealth ICBM wouldn't achieve as much as one would naively assume, and boost-phase detection would still invite retaliation.
I bet that given sufficient satellite coverage and image processing, ICBMs might remain visible throughout most of their flight, making radar stealth a moot point. I hope the world will not go too far down the road of building out measures and countermeasures to update our nuclear weapons systems, but I think you'd have to add that possibility to the long list of possible countermeasures when doing your planning.
With the potential for spacelaunch to become drastically less expensive due to reusable launch vehicles, a constellation of manned hypersonic glide vehicles may become economically practical.
The Titan II GLV was a human-rated adaption of an ICBM. Conceivably you could even put people in a truck launch ICBM.