The theory used to compute MTBF's assumes that all failures are spread
out over time by some distribution function. (Often the distribution
function is flat). And that failures are uncorrelated.

But now you see a correlated failure. One really vague possibility is
that the failure of one solenoid caused the other. More specific
examples of this could include one overheating next to another, causing
it to get too hot too.

Another possibility is that the same environmental condition common to
both solenoids caused both to fail. For example, vibration. Another
example might be a large voltage on the housing (is it metallic?)
holding both solenoids. Or a large voltage lighning-induced on the
wiring (which is after all nearly identical for both solenoids.)
Although you'd expect lightning to take out the drivers too.

Finally, some devices that come in pairs are well-known for failing in
pairs. Car headlights, for example. They were manufactured
identically, installed identically, operated identically, so you
shouldn't be surprised when they fail at nearly the same time. This is
sort-of modern manufacturing coming back to bite you... (Of course part
of this goes back to the possibility of an common extreme environmental
condition causing the fault - vibration, overvoltage, etc.)

Tim.