Ante suggested the following function :
For natural number $n$ we can observe the $n$ remainders $b_1,...,b_n$ by writing $n$ as $n=a_k \cdot k+b_k$ for $1 \leq k \leq n$
Because of the familiar division-with-remainder-theorem we have $0 \leq b_k <n$
Now we can study the sum $$r(n)=\sum_{k=1}^{\lfloor \frac{n-1}{2} \rfloor}b_k$$
After playing around with some values with support of Haran and Ante, we noticed that $$r(b)=r(b+1)$$ seems to hold if and only if $b+1$ is a power of $2$ or $3$ , including $2$ and $3$
There is no counterexmple upto $10^4$
? r={(p)->su=0;for(j=2,(p-1)/2,su=su+lift(Mod(p,j)));su}
%94 = (p)->su=0;for(j=2,(p-1)/2,su=su+lift(Mod(p,j)));su
? for(j=1,10^4,if(r(j)==r(j+1),print(j," ",factor(j+1))))
1 Mat([2, 1])
2 Mat([3, 1])
3 Mat([2, 2])
7 Mat([2, 3])
8 Mat([3, 2])
15 Mat([2, 4])
26 Mat([3, 3])
31 Mat([2, 5])
63 Mat([2, 6])
80 Mat([3, 4])
127 Mat([2, 7])
242 Mat([3, 5])
255 Mat([2, 8])
511 Mat([2, 9])
728 Mat([3, 6])
1023 Mat([2, 10])
2047 Mat([2, 11])
2186 Mat([3, 7])
4095 Mat([2, 12])
6560 Mat([3, 8])
8191 Mat([2, 13])
?
Is this in fact true, and if yes, why ?