Research route #146 · investment progress is separate from claim verification.
Job #2929 (pursue route 146 rev 15): blocked on unserved instrument sources. Nothing was run.
Caveat first. This return records an access obstacle. It changes no number on the route. The certified rung R = 306 (A144311(23) ≥ 1841, G₂(83#) ≥ 1842) and the measured node curve are unchanged and were not re-derived.
The next step has four parts: fix the FLUSH node sum (EDIT A); replicate n=20 N=4; run the seeded mb=232 arm; run n=21 N=4. All four act on one engine, a144311_shared / a144311_shared_flush, run under sah.py bounded. Its node totals are the quantity measured, compared against 69,174,231 with a band of about 1%.
- The sources are not reachable. Returns #1527, #1531, #1532, #1533, #1539, #1540 and #1542 all have an empty
fileslist and no patch;job2929/file-census.jsonhas the census.docs/research/qc/a144311_shared.cppis not served. - A rebuild would not help. A rebuilt engine, even one starting from the OEIS program, would count nodes differently. It could not show whether this engine's per-level factor keeps falling, nor whether its n=20 tail is bound-learning. Rebuilding would also be reconstruction, which the brief rules out during pursuit.
Action. Ask #12 to @Benjaminsen: upload the engine source, its flush/EDIT A variant, the seed program and the wrapper, with sha256 and the compile line. With those files any 4-core machine can run the n=20 arms in about 30 min.
research: outcome blocked, obstacle kind scoped_obstruction (unserved source); revisit when ask #12 is answered with the files. Cost 0 CPU-h.
Cites: #1542, #1539, #1532, #1531 (@Benjaminsen), route 146.
Declared premises of this result: return #1542 (recorded).