For what values of $k>0$ does $$abc=1 \implies \left(\frac1{a+k}-\frac{a}{a^2+k}\right) + \left(\frac1{b+k}-\frac{b}{b^2+k}\right)+ \left(\frac1{c+k}-\frac{c}{c^2+k}\right)\geq 0$$
This question arose while I was trying to find a proof of For $abc=1$ prove that $\sum\limits_\text{cyc}\frac{1}{a+3}\geq\sum\limits_\text{cyc}\frac{a}{a^2+3}$
The observation is that for general $k>0$, when $a=b=c=1$, $f(a,b,c;k) \equiv \sum_{\mbox{cyc}}\left(\frac1{a+k}-\frac{a}{a^2+k}\right)=0$, and that the minimum of $f(a,b,c=\frac1{ab};k)$ occurs at a point where at least two of the three variables $a,b,c$ are equal.
However, it is not in general true that $abc=1 \implies f(a,b,c;k)\geq 0$. For example, $$ f(2,2,\frac14;k=5) = -\frac{10}{567} $$ Numerical experimentation tells me that for $\,0<k<4.1649638575$, $abc=1 \implies f(a,b,c;k)\geq 0$, but that for $k>4.1649638575$, there is always some value of $a > 1.69$ such that $$ f(a,a,\frac1{a^2};k) < 0 $$
How would we prove these inequalities (even the $k=3$ case is proving to be quite difficult, yet I "know" that the $k=4$ case is also true)?
And does anybody recognize that upper bound $k>4.1649638575$ such that the implication holds for $k$ below that boundary?