get_random_realistic_address returns the same address for each zip

[bmos]

To reproduce:
python3 -m generate_fake_list --real-address-zips "04101"

The issue is that

geocoder.forward("04101", country=["us"])

returns a single result:

{
  'type': 'FeatureCollection',
  'query': ['04101'],
  'features': [
    {
      'id': 'postcode.9441004',
      'type': 'Feature',
      'place_type': ['postcode'],
      'relevance': 1,
      'properties': {'mapbox_id': 'dXJuOm1ieHBsYzprQTdz'},
      'text': '04101',
      'place_name': 'Portland, Maine 04101, United States',
      'bbox': [-70.284472, 43.6427499, -70.1815676, 43.675744],
      'center': [-70.25486, 43.661028],
      'geometry': {
        'type': 'Point',
        'coordinates': [-70.25486, 43.661028]
      },
      'context': [
        {'id': 'place.265562348', 'mapbox_id': 'dXJuOm1ieHBsYzpEOVFvN0E', 'wikidata': 'Q49201', 'text': 'Portland'},
        {'id': 'district.5670636', 'mapbox_id': 'dXJuOm1ieHBsYzpWb2Jz', 'wikidata': 'Q497788', 'text': 'Cumberland County'},
        {'id': 'region.337132', 'mapbox_id': 'dXJuOm1ieHBsYzpCU1Rz', 'wikidata': 'Q724', 'short_code': 'US-ME', 'text': 'Maine'},
        {'id': 'country.8940', 'mapbox_id': 'dXJuOm1ieHBsYzpJdXc', 'wikidata': 'Q30', 'short_code': 'us', 'text': 'United States'}
      ]
    }
  ]
}

[bmos]

so far, I have not found any good way to select random real addresses for a zip code.

The only possible solution seems to be using a fixed dataset like https://openaddresses.io/.
But this still will not work very well in covering some areas.