With `disable_harq == 1` set in the gNB config, the BLER-driven MCS adaptation mechanism is silently inoperative. `bler_stats->bler` never moves from its initialization value `(bler_options->lower + bler_options->upper) / 2` regardless of actual link quality. There is no fall-back AMC, so the gNB has no ability to react to changing channel conditions. Although disabling HARQ is intentional for NTN GEO scenarios, this does not imply that AMC should be disabled. ### Reproduction 1. gNB configured as GEO satellites NTN tutorial with `disable_harq = 1`. 2. Attach a UE and generate traffic. 3. Observe the periodic MAC stats: `BLER 0.10000` and constant MCS. A fix addressing this is on branch [amc-ntn-geo](https://gitlab.eurecom.fr/oai/openairinterface5g/-/commits/amc-ntn-geo). The fix (a) switches the BLER metric in `get_mcs_from_bler()` from `stats->rounds[1]` to `stats->errors`, and (b) removes the `harq_round_max == 1` short-circuits in `pf_dl()` (DL) and `pf_ul()` (UL) so the OLLA runs in all regimes. Tested with `2026.w19`, a USRP B210 as gNB, a USRP B210 as UE and a Propsim FS16 as satellite channel emulator.