Why, exactly?
Consider this: there has been ONE ballistic missile that was test-fired with a live nuclear warhead. That was the original Polaris A-1. Even that test was loaded, as all ballistic missile tests are. (The missile was not fired on anything resembling an operational trajectory, and it was extensively checked out prior to launch.)
That missile was retired from service shortly before I was born.
Now, fast-forward 41 years from that test, and 38 years from that missile's retirement.
Not one missile in the Russian or the US arsenal has been live-fired. They have undergone test launches. Those test launches feature extensive maintenance and testing from the missile manufacturer for several weeks prior to launch. After all that work by extremely knowledgeable engineers and technicians, the reliability of the missiles in these tests are roughly those of each country's space launch platforms (Delta, Proton, et cetera).
The result: after extensive maintenance and checkout that would not be performed prior to an attack (due to a shortage of skilled technicians and because extensive maintenance work would be easily noticed prior to an attack), a US missile will fail to reach its target 20% of the time. A Russian missile will fail to reach its target 25% of the time.
Combat firings would be very different. The failure rates would be much higher (probably double, at a minimum, the "test launch" failure rate). Bottom line: the Russians are looking at a failure rate of at least 50%, if not higher.
Now, look at the overall force structure of the US and Russian nuclear forces. A large difference jumps out at you.
The US force includes a significant bomber component. Bombers and cruise missiles have much higher demonstrated reliability compared to their missile brethren. The bombers are also reusable.
Additionally, note the historical performance of the Russian air defense forces. In 1978, a Korean Air Lines flight got very lost. After a comedy of errors, the Soviet PVO (Troops of Air Defense) managed to shoot down the bomber near Finland.
Heads rolled. Surely, after those heads rolled, things got straightened out, right?
Wrong.
Five years later, another KAL flight flew into Russian airspace off Kamchatka. After an even longer-running and more error-filled sequence of events, the plane was shot down.
Again, heads rolled.
And then, in 1987, a German teenager landed a Cessna in Red Square. He wasn't ever fired on. Hell, he wasn't even detected until somebody looked out a window in the Kremlin.
Well, at least heads rolled. But heads rolling is no guarantee of future performance...and there has been no test of the system since then.
Note that the circumstances that existed in each of these cases. There was no defense suppression effort accompanying these aircraft--but the PVO damn near let two of them escape, and the third one could've dropped off a W80 into the Kremlin courtyard as a candygram before anyone knew it was there. In a full-on war, with jamming, defense suppression strikes, and many more targets, the PVO is most likely not going to cover itself in glory.
And, finally, US bombers can deliver unguided bombs to within FEET of their intended impact points.
All of these points are seriously considered by the Russian leadership.
A full Russian strike would likely do far less damage to the US than they wished or deemed necessary for victory; it would do enough damage to the US to whip the American public into a genocidal fury; and the American's retaliation, relying as much as it does on bombers, would likely be able to reach out and strike the Russian leadership.
When uncorking the genie's bottle is likely to result in your own death...what is the incentive for doing so?