1.401(a)(4)-3(B)(5) contains a safe harbor for insurance contract (412(i)) plans.

1.401(a)(4)-3(B)(5)(iii) says that the benefit formula must be one that would satisfy 1.401(a)(4)-3(B)(4) if the normal retirement benefit were accrued ratably over each employee's period of plan participation.

1.401(a)(4)-3(B)(4)(i)©(2) says that the normal retirement benefit must be a flat benefit that requires a minimum of 25 years of service at normal retirement age for an employee to receive the unreduced flat benefit.

So 1.401(a)(4)-3(B)(5)(iii) seems to indicate that for a 412(i) plan, the benefit formula must be a flat benefit that requires a minimum of 25 years of plan participation at normal retirement age for an employee to receive the unreduced flat benefit.

Do I have this correct or is it just that a 412(i) plan must have a flat benefit that accrues ratably over an employee's years of plan participation with no minimum number of years of participation required to get the full flat benefit at retirement?

I think the key phrase here is "safe harbor". If you were setting up a 412(i) for a "one-man" plan, your concerns of safe harbor formulas don't matter.

The purpose of a 401(a)(4) safe harbor is to provide guidance regarding how to design a plan such that it can avoid testing for nondiscrimination in contributions or benefits.

In a one-man plan, discrimination is impossible.

So I'm pretty sure that a safe harbor anticipates that there are NHCE employee participants since it tells us how to avoid discriminating [in the context of 401(a)(4)] against them.

Does anyone have an answer to my question assuming that the 412(i) plan has at least one non-owner, NHCE participant?

Actually, wasn't trying to avoid your question. But the key thing is that if you want a "safe harbor" 412i plan, then your formula has to conform to the 25-year rule on participation. This doesn't mean that you can't have a differing methodology of benefit formula in a 412i plan, but if you do it isn't a safe harbor and you'll have to do the General Test.