This is how I went about it.
The crux step is this: If for any t_1, t_2 \in \mathbb{R}^+ we can show that f(t_1) = f(t_2), we can conclude that f(x) is a constant function.
So we argue that given t_1, t_2 \in \mathbb{R}^+ we can find x,y \in \mathbb{R}^+ such that x+y = t_1; x^2+y^2 = t_2 and therefore from the problem statement f(t_1) =f(t_2)
But we have the inequality 2(x^2+y^2) \ge (x+y)^2 and so 2t_1 \ge t_2^2 must hold for us to directly use the method outlined above.
In case this does happen, then, to guarantee the existence of such x,y, a geometric argument will suffice. x^2+y^2 = t_2 is a circle and x+y=t_1 is a line and the intersection of these two will take place in the first quadrant, and so we have our (x,y)
Three cases therefore need to be sorted out:
1) 2<t_1; 0 < t_2 \le 2
Its obvious that 2t_1 \ge t_2^2 holds
2) 0 < t_1,t_2 <2
I will argue that either 2t_1 \ge t_2^2 or 2t_2 \ge t_1^2.
If to the contrary, we have both 2t_1 < t_2^2 and 2t_2 < t_1^2, then multiplying, we obtain t_1t_2>4 which is not possible as 0 < t_1,t_2 <2
3) 2<t_1, t_2
Now there is no guarantee that either of the inequalities will hold. But, we use here the result obtained by some of you that f(2t) = f(2t^2) or written in another way f(t) = f\left(\frac{t^2}{2} \right)
When t>2, the sequence t_{n} = \frac{t_{n-1}^2}{2} is a strictly increasing sequence
This tells us that we can substitute t1 with a large enough t to satisfy 2t \ge t_2^2
Thus all possible cases have been disposed off.
Hence, to summarise, given any t_1, t_2 \in \mathbb{R}^+, we can find equivalently t_1', t_2' \in \mathbb{R}^+ such that
1) f(t_1) = f(t_1'); f(t_2) = f(t_2')
2) \exists \ x,y \in \mathbb{R}^+ with x+y = t_1'; x^2+y^2 = t_2'
so that we may conclude that f(t_1) = f(t_2)
Since this is true for any arbitrary t1, t2, we must have f as the constant function