Cap regional failure oversampling

docs/BELGIUM_BUILDING_TRAINING_LOOP.md

@@ -108,6 +108,11 @@ modes without copying a protected AOI into training. The sampling evidence
 records both context sets and repeat factors. When no matching train context
 exists, regional sampling remains active and the missing context becomes a
 concrete input for the next immutable corpus expansion.
+Failure weighting may not let one region exceed 65% of the sampled entries.
+The deterministic cap removes only repeated entries and retains every unique
+train tile at least once; manifests record pre-cap counts, final counts and the
+number of dropped repeats. This keeps a weak region prominent without turning
+the national detector into a single-region expert.
 The checkpointed orchestrator invokes this builder after every rejected
 iteration, stores its checksum in `training-loop-state.json`, and uses the
 resulting dataset YAML for the next checkpoint. A restart resumes both the

docs/CODEX_EXECUTION_LOG.md

@@ -15,6 +15,10 @@
   loop now requires and hashes the train tile-quality report, and refuses
   missing counters instead of treating absent invalid/blank-label evidence as
   zero.
+- Capped failure-driven regional oversampling at 65% after the first v31
+  sampling assigned 75.5% of entries to Flanders. The cap retains every unique
+  tile, removes repeats only and writes pre/post regional counts into the
+  checksummed sampling evidence.
 
 ## 2026-07-27 - Guest demo and product professionalization