When will it happen again? - Models
Scientists cannot predict precisely when the next 1906-like earthquake will happen. Unfortunately, earthquakes do not produce known warning signs just before they occur. So estimates of when a large earthquake will occur are imprecise and are based on "models."
A model is a simplified idea of how something works. For the repeat of the 1906 earthquake, the best models consist of two parts: the THEORY of plate motion and the accumulation of stress along a locked fault, and OBSERVATIONS of past earthquakes on that fault and the rate the plates are now moving. These models suggest that it might take 200 years or more (starting in 1906) before enough stress accumulates on the fault to produce another great earthquake. (The long-term rate of motion, averaged over many earthquake cycles, on the 1906-segment of the San Andreas fault is between 3/4 to 1 inch per year. At this rate, a 20 ft earthquake offset requires 200-240 years to accumulate.)
But this estimate is not precise because earthquakes do not occur "like clockwork", and because other factors not included in our simple model may be involved. So, while the most likely time for a 1906-like earthquake to strike again is perhaps late in the next century, there is a small chance (about 2 percent) that it could happen in the next 30 years.
However, this does not mean that the San Francisco Bay region will not experience damaging earthquakes in the next hundred years.
Models for the Hayward fault, the Rogers Creek fault and the Peninsula section of the San Andreas fault suggest that smaller earthquakes (say, ones with magnitude 7) are also expected, and are now much more likely to cause serious damage than a repeat of the Magnitude 8, 1906 earthquake.
Shaking is just as intense in an M7 as it is in an M8, but it doesn't last as long and isn't as widespread. Main danger now to Bay area is from M7's.
If the 19th century is an example, then we can expect one or more M7 earthquakes on Bay area faults in the decades leading up to the repeat of the 1906 earthquake -- i.e., over the next several decades.
The Working Group on California Earthquake Probabilities estimated after the 1989 Loma Prieta earthquake that the probability of a magnitude 7 earthquake in the next 30 years in the San Francisco Bay region was 67 percent. (U.S.G.S. Circular 1053)