Across 300 sampled payloads from today, `eta_c_estimate` stayed pinned at the 3.2 scan ceiling with zero or near-zero bracket width, including deep 2048 runs and smaller 256 cases. The soft mode was very symmetric between chambers, with `last_chamber_balance` usually about 0.99 to 1.00, and it was spectrally well separated, with `last_gap_ratio` about 3.01 at 2048 and about 3.09 at 256. But the bridge never actually loaded: `last_dike_mass` stayed only about 0.00350 at 2048 and only about 0.0287 at 256 against the 0.14 target, while both chamber masses also remained tiny, so the mode stayed overwhelmingly exterior rather than becoming a chamber-dike-chamber bridge.