There's a distinction between calculated MTBF and the statistically acquired MTBF that manufacturers (and only they) can provide. The mechanisms by which MTBF is calculated are often antiquated, too, being based on parameters no longer in use.

I do recall one aerospace guy calculating the MTBF of a board we were producing based on values applicable to 4000-series CMOS, commonly used in aerospace hardware, albeit a couple of generations earlier, when nearly everything was either LSTTL or HCMOS, neither of which benefitted from a 12-volt supply. Strangely enough, the government spec that he was required to use still relied on the old values assigned to logic components that did operate at up to 15 volts Vdd.

The distinction between calculated and statistical that I was making was between the JEDEC/mil-standard qualification process statistics, really only done by component makers, as opposed to the subsequent calculations, based on those statistics, made in the field by the reliability guys.

RE