TITLE: Uniformity of the distribution of the prime numbers on the prime residue classes (mod $m$)
QUESTION [6 upvotes]: Given positive integers $m$, $r$ and $n$, let $\pi(m,r,n)$ denote the number
of prime numbers $p \leq n$ in the residue class $r$ (mod $m$).
Further let $1 = r_1 < r_2 < \dots < r_{\varphi(m)} = m-1$ be the prime residues
(mod $m$), and let $\sigma \in {\rm Sym}(\{r_1, \dots, r_{\varphi(m)}\})$
be a permutation of them. Are there always infinitely many
$n \in \mathbb{N}$ such that
$\pi(m,\sigma(r_1),n) < \pi(m,\sigma(r_2),n) < \dots < \pi(m,\sigma(r_{\varphi(m)}),n)$?
If the answer is yes, are explicit bounds $B(m)$ known such that
for every permutation $\sigma$ of the prime residues (mod $m$) there is
at least one such $n$ less than $B(m)$?
If the answer is no, for which permutations $\sigma$ are there infinitely many
such $n$, and for which are not?

REPLY [12 votes]: The answer is conjectured to be "yes." Further, assuming GRH and GSH Rubinstein and Sarnak, Chebyshev's Bias, Exp. Math 1994 even could show results on the logarithmic densities of the resepactive sets (which in particular is non-zero). I do not know about general explicit upper bounds.
Without GRH a lot less is known. The problem you mention sometimes goes by Shanks--Rényi race problem. A very good exposition of this circle of ideas is Prime Number Races by Granville and Martin; see page 22 and subsequent ones for the discussion of Rubinstein and Sarnak's results. (Added note: this is in fact the same paper Benjamin Dickman mentions in a comment I had not seen before.)