Remarks on a problem in the ``Elemente'' (Q1226514)

Let \(p\) denote an odd prime and \(\ell(p)\) the order of \(2 \bmod p\). Write \(E(r)\) for the number of odd primes \(p\) with \(\ell(p)=r\) and \(A(x,\delta)\) for the number of odd primes \(p\) with \(p\leq x\) and \(\ell(p)>p^\delta\), \(0< \delta <1\). \textit{G. Jaeschke} [Aufgabe 618 in Elem. Math. 26, 43 (1971) solved by \textit{P. Bundschuh}, hence the uniformative title of the present paper] proved that \[ E(r) \leq {r \log 2 \over \log r}, \tag{1} \] \[ A(x, \delta)=(1+o(1))\frac{x}{\log x}, \quad 0< \delta <\tfrac 12, \tag{ii} \] \[ A(x,\tfrac 12) \geq (1- \log 2+o(1))\frac{x}{\log x}. \tag{iii} \] In a few lines, using the Chinese remainder theorem and a weak form of Stirling's formula, (i) is sharpened to \[ E(r) \leq ({\tfrac 12}+o(1))\frac{r\log 2}{\log r}. \] The main result of the paper is that (ii) also holds for \(\delta =\tfrac 12\). The proof of this is rather complicated and uses among other things an estimate from the sieve method of Brun. There are some conjectures on \(E(r)\) and \(A(x, \delta)\) which seem very difficult to prove. To mention one of them: \[ E(r)=o(r^\varepsilon),\quad \varepsilon >0. \]