MIL-217 Weakest link philosophy for SMD

[741]

I don't understand the MIL justification for using 'weakest link'. (MIL-HDBK-217F, NOTICE 2, Section 16.2).

Why is it that once package thermal expansion is factored in, only the 'weakest link' is considered - surely we still have a simple MTBF figure, all we have done is make it more accurate?
For instance if there are 2 SMT devices with the same package, and they (both) share the "weakest link" attribute, then how can the existence of the 2nd device be ignored?
As I see it, after accounting for thermal cycling, the MTBF of both devices would then add in the usual way, giving a new MTBF that is half that for one device.
Further, I also wonder how we can ignore other SMT devices, even those with somewhat lower 'failure rate'.

(Note: I'm new to the whole MIL-217 thing, and apologies if this post is not sensible.)