Let me start by paraphrasing Richard Nixon by saying "I am not a crank!"; I am not "working on" Goldbach's Conjecture, this is an idle curiosity along similar lines to this question.
So, as any crank knows, you look at $2m$, and then you start looking at $2m-p$ for primes $p$. Well how about we place $2m$ in an interval where for about $k\approx N/\ln N$ the first $k$ of the $2m-p$ (i.e. $2m-3,\,2m-5,\,2m-7,\dots,\,2m-p_k$) are all composite.
We can do this readily by considering $2m$ in the interval $[N!+2,N!+N]$, say $2m\approx N!+N$.
So some focused questions. The first a cursory Google could not answer for me; the second I doubt Google could answer.
Has anyone seriously looked for counterexamples in such an interval? If not, why not?
I understand that, using this notation, all such intervals up $N=20$ have been accounted for via verification up to $\approx 5\times 10^{18}$.