EvW was a complete disaster from start to finish. A game is not a mod.
There's some real conceptual issues that need to be overcome to make a Cold War game fun.
Most of all nuclear war (which should last 24-48 hours tops) is going to be hard to make fun in a game that lasts 40-50 years. There is a real problem that nuclear war is both 1) something that shouldn't happen and 2) something the player will have to spend the entire game preparing for.
Just why should the player build nuclear weapons if the aim is not to use them? If the player is forced to build them by a mechanism (e.g., you have to keep parity with the enemy otherwise you lose the game) then in what sense is this fun and not simply microing a meaningless statistic?
CO-IN warfare is something that has, historically, not been fun in Paradox games. HOI4 dealt with this by abstracting it. You can't do this in a Cold War game as CO-IN warfare made up the majority of the fightiing in that period. The solution suggested by the EvW team (basically the HOI3 guerilla warfare model on steroids) would have been attrocious whack-a-mole - and unrealistic given just how rare division-v-division scale combat is in CO-IN wars.
I'm not saying it can't be done, but it doesn't look easy.
And thus realise what a pile of garbage it was, and how it would never, ever be completed in any real time-scale and if, by some miracle, it had been, it wouldn't have been any fun to play.
We can work from this though. The pitfall of EvW was the HoI tag on the title; they were making a Hearts of Iron game set in the Cold War, instead of straight up making a Cold War game. So rather than thinking "What can we do to make a good cold war game", they were thinking "What can we change to make Hearts of Iron into a cold war game".
So, rather than start from an existing feature base, we start from a series of goals. I'll do this for nuclear weapons and nuclear policies, as it is the simpler issue (relatively speaking).
Regarding Nuclear War, you identified our gameplay goals; the player must be incentivized to prepare for nuclear war, while being disincentivized to begin nuclear war. This can be done through a high-risk high-reward structure. We can also agree, as you said, that nuclear build-up should not be arbitrary. This can be achieved in X steps; 1st, make nuclear war devastating but not an arbitrary game over. Thus, players want to use it offensively and avoid it defensively. 2nd, teach the AI (and thus the player) MAD, Massive Retaliation, and the diplomacy tools to benefit. 3rd, add First Strike Capability theory, but make it difficult to acquire. 4th, add diplomatic and internal repercussions to even successful safe first strike.
1. Nuclear war should be devastating for the target, destroying military targets and disrupting the economy of the state and thus, the political stability of the state, highly raising dissent or lowering national unity or some other similar mechanic. Being struck by nuclear war should be devastating; not an immediate game-over however. A full nuclear exchange should cripple a nation, but a crippled nation can be saved in time. A limited nuclear exchange should harm a nation but not cripple it. Limited nuclear exchanges will be rare, because of a) MAD, and B) Massive Retaliation. Time to recover, however, is a valuable resource. This should be made clear to the players, that while they can technically survive nuclear strikes, it will in all likelihood drop them an entire tier of operation; from superpower to major power, from major power to minor power, from minor power to weak or failed state, due to the economic and political effects of nuclear strikes. This is your underlying motivation for all nuclear game-play; get the ability to use nuclear force, and get the ability to avoid others using it, because nuclear strikes are deciding factors in war.
2. So, Nuclear strikes are devastating, so all actors a) want to be able to use it, and b) want to be able to avoid it. The next step in having fun and useful Nuclear War, is to give them tools to avoid it. There are two aspects to this; MAD and Massive Retaliation. Technically, Second Strike Capabilities also fall under this section, but we'll talk about those in part 3. MAD is the main factor here, and it teaches the AI to respond to nuclear force with nuclear force, and vice versa, that any nuclear strike it carries out will be responded to with nuclear strikes. The AI will be trained and the players told to avoid being hit with nuclear strikes at (nearly) all costs. Thus, they will avoid striking other major nuclear powers. More minor nuclear powers, like India and Pakistan, may be able to have a nuclear exchange without total destruction, but will still end up crippling each other. Massive Retaliation is an extension of this theory, possibly one that can be toggled on and off. It is, simply, the concept that any force will be met with massive nuclear force. This removes the concept of limited nuclear exchanges between superpowers, so all nuclear exchanges will involve nearly all the capabilities of the involved powers. It also anchors the Cold War more generally in Europe, preventing, while the policy is activated by the superpowers, direct conventional war between major powers. Diplomatic tools allow superpowers to extend Massive Retaliation protection to non-superpowers, perhaps even to all conflicts, preventing the US from bombing Vietnam, and the USSR from bombing Afghanistan. This can also be done by making it hard to divine nuclear trajectories; if an ICBM is spotted, it might not be immediately clear who fired it and at whom, so Massive Retaliation kicks in. Teaching the AI to respond to any strike that MAY be headed towards them as if it IS would fix this.
3. So, we now have tools for the actors to avoid nuclear strikes, but we want the possibility for nuclear war to, in fact, happen, so we must teach the AI to value First Strike Capabilities, the ability to overcome the factors in point 2. If actors can gain the technological abilities to strike without being struck in return, then they must be aware of this, and be more likely to use those capabilities. Likewise, actors must fear others gaining First Strike Capabilities, and develop Second Strike Capabilities. A tech war between first and second strike capabilities and the abilities to circumvent them will dominate the late-game (as no serious first strike capabilities existed in the early cold war). The only truly safe first-strike capability would be defensive in nature, the ability to destroy every incoming missile, and would be the "holy grail" of the game. It would be expensive and difficult to acquire.
4. We don't want the decision to use nuclear force to be a given even if safe First Strike capability is acquired. Even in cases where a nuclear strike would not lead to a response, there should be a system in place to not make it the 100% best option to strike; This can be done by a) giving diplomatic penalties to those who use nuclear weapons, and b) giving internal dissent issues to those who use nuclear weapons, as their own citizens are disgusted by the horrors of nuclear war, even when inflicted on other people.
There. Now the threat of Nuclear War hangs over the entire game. The player wants to use nuclear weapons where they think they can get away with it, because the effects are devastating, but they don't have the ability to absorb nuclear strikes without severe long-term effects. The only way to avoid nuclear strikes is to use MAD, or to use the Massive Retaliation or Nuclear Umbrella policies of other nuclear states. First-Strike Capabilities threaten to up-end this, so players and their enemies race in technology to see who can get a definite advantage. Even relatively safe nuclear strikes will have diplomatic and internal dissent consequences, so that the decision to use nuclear force isn't an obvious one.
This is all done via AI learning recognizing specific situations. It is the job of the mechanics to create the situations to enable the 4 points above. I think that's toally doable.
