Training bases that don't have a significant active unit component (example Ft McCoy in Wisconsin) are going to be way down the target list. Depending upon how fast a major strategic exchange develops, units at such bases can leave to pre-assigned rally points where some gear etc has been stashed. An example of this would be rations and tentage stashed at a state or national park not too close to a given base, trucks ready to go etc which means that even a couple of hours lead time means you can get folks with light weapons and ammo there and out of the danger zone. They can then take a deep breath and proceed to a designated site where they can be utilized - there ought to be several potential "final" destinations as one or more may be eliminated.
Desertion at all levels will be a problem, basically the more a command structure is intact, with the local command element key although national authority is necessary. Scared recruits are one problem, another is men who have families they know or believe to have survived whatever exchange and think they have a shot at getting from where they are to where their family is. If there is decent unit integrity at the platoon/company level and then higher, this ought to be manageable. Small broken groups are likely to be the biggest issues as they may descend in to banditry/warlordism.
There are only going to be so many ICBMs/SLBMs ready to launch for the USSR at any moment. Some targets may need to be serviced by several to achieve the results. A training base can't hurt you right now, and even longer term is a minimal threat, therefore you won't waste a missile on one of these maybe a bomber may have this on their target list or it may be an alternate target for a bomber. In 1984 the USSR had roughly 1,400 land based intercontinental missiles. These have a total of roughly 6,000 warheads with ~300 missiles have 10 MIRVs and ~350 having 6 MIRVs. There are roughly 60 SSBNs of all sorts with a total of ~920 missiles some of which have 3-5 MIRVs - some of these can operate effectively from "bastions" close to the USSR, others require deployment in to the Atlantic. A small percentage of the land based missiles are mobile.
Some land based missiles, especially the mobile ones, would be held back for second strikes for targets that were not serviced in the first strike and also as power chips after the war. Likewise for some of the SLBMs. So now the math starts. Assume there are 8,000 warheads than can be delivered by missiles. Assume between mobile ICBMs and SLBMs held back you now have a theoretical max of 6,000 warheads deliverable by missiles. Starting with the SSBNs, no matter how hard the Soviets (or the Americans) try some of them will be in port if it hits the fan, and at some point the SSBNs will be targeted by the other side even if left alone while things are conventional and atoms are not on the horizon. The US and UK have very capable ASW forces, so there will be Soviet losses among deployed boomers. For the silo based missiles, no matter what a certain percentage are simply not going to be ready to go when the button is pushed. If the launch is delayed for some reason, deliberate or just circumstances, the missile may die in its silo. For all missiles a certain number will malfunction so that they don't reach the USA, a certain number of MIRVs will malfunction and perhaps not separate from the bus. Some of the warheads will just go splat, and not explode and if some of the conventional explosives in the device do go off simply spread a relatively small area of plutonium goodness. So the question is how many warheads actually reach the USA and land reasonably close to their targets (remember CEP says 50% of the missiles hit in that circle, the other 50% from 1 cm outside the circle to anywhere on the planet).
Given the estimates of max readiness for the sub force and the missile force, failure rates for the various missiles, and how many would be held for second strike or later you can estimate the max number of warheads that might hit the USA. Of course some will be used against Canada, some against various US bases outside CONUS etc. It won't be a small number but it will be substantially less than the theoretical max of 8,000 if all fired at the US/US owned targets. Training bases like the Marine boot camp in San Diego as well as the Navy boot camp there will go, simply because San Diego is going to be hit by multiple warheads and they will go as "bonus". Will Parris Island go? Probably not by missile, no active forces there. Will the Navy boot camp at Great Lakes go - it may get caught in stuff hitting Chicago, but it is potentially far enough away to survive depending on impact points. There are many other basic training bases and others that might get lucky and be low down on the list. Don't forget targets due to be serviced by missiles that fail, or on the list for SSBNs that are sunk before they fire will escape unless and until the USSR knows they were missed, wants to kill them, and has the capability to do so. Any target slated for only one missile warhead, be it a military installation or a city can be the beneficiary of missile failure, sub sinking or just poor guidance.
Any target not hit by a missile in the initial salvo is probably safe from missile attack for some time and possibly permanently. If they are going to be hit by a bomber, this means the targets have several hours before they hit the vulnerable zone. Assuming they have plans to evacuate some or most of their assets somewhere not likely to be on a target list (and one hopes these plans would be made) once the nuclear exchange starts they have more than enough to to bug out.