As it happens, the chance of a misfire on any motor in a cluster is very slightly positively correlated with that on any other motor in the cluster. If we assume that the chance of any motor misfiring is independent, therefore, our conclusions should be very slightly optimistic. Let's try that.
Let the chance of a misfire be represented as P, so that the chance of a successful ignition for any motor is (1-P). The chance of N motors igniting successfully, under the assumption of independence, is (1-P)^N.
Here are the chances of successful ignition in clusters of sizes 1-20, given (1-P)=0.99 and (1-P)=0.95. Again, these are slightly optimistic. (*** added just to space the columns)
You can get an order of magnitude handle on P by your experience with single motor launches. That's not a perfect estimate (You might take more care on a cluster and use different igniters. OTOH, single ignitions don't have a strict time limit as cluster igniotions do.). I suggest the comparison just to demonstrate that (1-P) is not unity or extremely close to it.
Most N's would likely be small, and three or four consecutive successes are quite likely. There's still that chance...
HYPOTHETICAL SUCCESS CHANCE TABLE
N
1 0.990000000 ***0.95
2 0.980100000 ***0.9025
3 0.970299000 ***0.857375
4 0.960596010 ***0.81450625
5 0.950990050 ***0.773780938
6 0.941480149 ***0.735091891
7 0.932065348 ***0.698337296
8 0.922744694 ***0.663420431
9 0.913517247 ***0.63024941
10 0.904382075 ***0.598736939
11 0.895338254 ***0.568800092
12 0.886384872 ***0.540360088
13 0.877521023 ***0.513342083
14 0.868745813 ***0.487674979
15 0.860058355 ***0.46329123
16 0.851457771 ***0.440126669
17 0.842943193 ***0.418120335
18 0.834513761 ***0.397214318
19 0.826168624 ***0.377353603
20 0.817906938 ***0.358485922
Luck and Regards,
-LarryC