The commander of the Strategic Rocket Forces in an interview published last month disclosed an interesting number - during its 45-year history the rocket forces conducted 5136 missile launches, of which 8% were failures. More that a third of all launches was conducted for training purposes. Note that the number includes launches of medium-range missiles (like SS-20, R-12 and R-14) as well as those of ICBMs, for they were also in the Strategic Rocket Forces.
92 per cent does not look like a high figure for reliability of missiles. However, keep in mind that a number of launches were conducted during development, when one would expect a higher failure rate (and it seems that the Soviet designers used a lot of missiles during development). For missiles accepted for service reliability seems to be higher - the exact numbers are hard to come by, but when the Soviet Union was liquidating somoe of its SS-20 missiles by launching them, all 72 launches were successful. This indicates very high reliability (without getting into statistical details, at least 99%). Another hint on missile reliability is the two salvo launches of R-39 (SS-N-20) missiles, conducted in mid-90s. Of 40 missiles (two Typhoon-submarine loads) only one got stuck in a silo, idicating the overall system reliability of about 98% (but the missile itself would have probably work if it managed to leave its silo).
There were, of course, recent failures as well - two attempts to launch a R-29RM (SS-N-23) missile from the Novomoskovsk submarine of the Project 667BDRM class in February were unsuccessful because of a failure of the control equipment. A launch from a similar submarine the next day also ended in failure - this time the missile had to be destroyed shortly after launch due to guidance malfunction. This indicates that reliability of the D-9RM missile complex with R-29RM missiles is fairly low. However, it's hard to pin a meaningful number on it.