Let's consider ICBMs as a starting point. Re-entry time from 100kms is 2mins, and speed is typically 7km/s. Until they start re-entry, they're hard to track, as there's no rocket or re-entry heat. (Data from wikipedia)
100kms is obviously way lower than LEO. Two options are possible:
- They could be fired from LEO or a higher orbit. A number of other answers point to the high visibility of a retroburn. That may not be a huge issue if you don't know where they're going to land – and because they're hard to track until re-entry, it's not unlikely that the trajectory could remain hidden; I don't think a retroburn would necessarily be enough to make their trajectory known, especially if we try to obfuscate them; I think from an LEO even a fairly small change would have a substantial impact on landing point. Also, just because current reentry methods are very visible, doesn't mean we can't develop more subtle ones. There's never been any need to do so, so I doubt much research has been done. A civilisation otherwise capable of space travel would likely have other options.
- They could be dropped from high-altitude space-ships, or even deployed surface-surface via a missile system.
In either case, whilst in space, they're pretty safe, as they're hard to spot.
We've now got the issue of breaking something from ~7km/s.
Wikipedia gives the max acceleration for a human on a rocket sled at 46g. (wow!) Let's take that, and assume that a civilisation otherwise capable of droppods can find a way to reduce the effects of deceleration. We're thus looking at decelerating from 7km/s: 46g = 7000m/s / 152s – so two and a half minutes of breaking at max deceleration. It's gonna have to be rockets; parachutes won't provide that deceleration.
Lithobraking will be Lithobreaking, unless we've got some handwavium device to reduce the effects of deceleration – or we're dropping from substantially lower.
But, hold on a minute, why are we assuming we're dropping humans, and not combat robots? Suddenly, the deceleration issue largely disappears. We can decelerate a robot very fast (though again, probably not pure lithobraking – very little will survive that kind of impact).
How interceptable are these? Hmm... A stinger can be ignored; if that worked, they wouldn't have had to develop THAAD. We can assume THAAD isn't anywhere near 100% effective, as if it were, we'd not care about NK's missiles. Also, it's very expensive per shot.
So if we drop some empty pods or similar, we can easily make it too expensive to shoot them all down. Also, how many anti-drop-pod missiles would be available to fire concurrently? We can likely just drop far more pods than can be shot down at once.
So survivability is likely to not be an issue.
Back to the original question: are they militarily viable?
Currently, the vast quantity of rocket fuel they'd require would make each one extremely expensive, but a civilisation capable of space flight has presumably overcome the fuel cost.
They'd not be stealthy, but given the speed, how quickly a target would realise they were the target and respond is debatable.
It's unlikely they'd be fast-response, due to the time required to plan for their launch and be in the right place in orbit to launch (unless they're deployed by some kind of missile?)
TL;DR:
So we're looking scenarios which require a planned, but very fast, deployment. Potentially preceded a few seconds earlier by a bombardment (which also might intercept interceptor missiles). Great for any kind of attempt to secure a logistics point (bridge, transmitter, etc) or other asset which wouldn't respond meaningfully in the max ~2 mins warning they'd get. Probably not so great for a coup d'etat, as they're more likely to realise they're being targeted and respond, but still a viable option.
If we're dropping combat robots rather than humans, then much more viable.
