I am not asking about the fairly well-known effect of the earth "appearing less curved to radio waves" that are otherwise still essentially line-of-sight, but a deeper arcanum:
In the ARRL Antenna Book, 17th edition (1994) there is a discussion of "Reliable VHF coverage" in starting on page 23-7 in the Radio Wave Propagation chapter. The claim is made,
Because of age-old ideas, misconceptions about the coverage obtainable in our VHF bands persist. This reflects the thoughts that VHF waves travel only in straight lines, […] However, let us survey the picture in the light of modern wave-propagation knowledge and see what the bands above 50 MHz are good fro on a day-to-day basis, ignoring the anomalies [presumably referring to the tropospheric ducting of previous section] that may result in extensions of normal coverage.
It goes on, after mentioning an article by D.W. Bray, K2LMG in the November 1961 QST magazine, to present two graphs that plot "tropospheric path loss" against distance. The curves therein rise steeply from 120 dB of loss at a distance of 0 miles [?!] to around 180 dB near 50 miles, then level off slightly so that at 500 miles there is a path loss around 240 dB. (That's reading the 50% reliability chart roughly, there's actually 4 lines plotted for 144/50, 220, 432, and 1296 MHz, as well as a second separate chart showing 99% reliability; the 99% reliability chart is very approximately 10–20 dB worse than the 50% one at any given point.)
UPDATE: thanks to W0BTU Mike, here's the actual charts scanned from an earlier edition:
What "modern wave-propagation knowledge" is this referring to? What mechanism(s) would allow VHF signals to be 99% reliably received 500 miles away, albeit with more than 250 dB of path loss, or 50%-of-the-time reliability with a little less loss? (These path-loss charts do NOT assume any antenna-height gain.)