The site's wording defines $h_2(G)$ and then asks about $h(G)$, which it never defines, so the question is undefined for every graph. The change replaces "$h(G)$" in the question by "$h_2(G)$". The evidence is the posers' own text: Erdős, Gyárfás and Ruszinkó [EGR98], p. 493, let $h(G)$ be the least number of edges whose addition gives a maximal triangle-free extension, that is a triangle-free graph on the same vertex set of diameter at most two, and write "$h_2(G)=h(G)$"; their Problem 4.1 (pp. 498--499) asks the question in terms of $h(G)$. So the site's $h$ is the posers' name for the site's $h_2$, and the defect is the site's: it renamed the function in the definition but not in the question. The site's "diameter $2$" differs from the source's "diameter at most two" but changes nothing: for $n\geq3$ a triangle-free completion of diameter at most two cannot have diameter one, since it would then be a complete graph containing a triangle, and the question is asymptotic in $n$. No result about the site's wording is recorded.