specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Fair Housing and Equal Opportunity
providerId: fair-housing-and-equal-opportunity
generated: '2026-09-07'
method: probed
source: >-
  Response headers observed on a live unauthenticated GET to
  https://services.arcgis.com/VTyQ9soqVukalItT/arcgis/rest/services/Fair_Housing_Assistance_Program_Grantees/FeatureServer/0/query
  on 2026-09-07, plus the maxRecordCount values in the service metadata saved in
  arcgis/.
created: '2026-05-04'
modified: '2026-09-07'
supersedes: >-
  REPLACES a 2026-05-04 bulk-sweep scaffold that invented free / professional /
  enterprise request tiers and X-RateLimit-* headers this provider does not
  return. That was fabrication and has been removed.
limit_count: 0
limits: []
published_limits:
  documented: false
  note: >-
    FHEO and HUD document no rate limit for any of these surfaces. There is no
    published requests-per-second, no quota, and no throttling policy. An honest
    zero: nothing is published, so nothing is recorded.
headers:
  standard: false
  limit: null
  remaining: null
  reset: null
  retryAfter: null
  policy: null
observed_headers:
  - name: x-esri-org-request-units-per-min
    observed_value: usage=275;max=6000
    meaning: >-
      A live, per-minute request-unit budget for the ArcGIS Online organisation
      hosting these layers. This is the ONLY quantitative capacity signal on the
      wire. It is Esri platform telemetry, not an FHEO-published limit, and the
      usage figure reflects ALL traffic to HUD's ArcGIS org, not just the
      caller's.
  - name: x-esri-query-request-units
    observed_value: '3'
    meaning: The cost in request units of the single query that was issued.
  - name: x-esri-tiles-basic-query-mode
    observed_value: 'true'
  - name: cache-control
    observed_value: public, max-age=30, s-maxage=30
    meaning: >-
      Responses are publicly cacheable for 30 seconds — the practical way to
      stay well inside the org budget.
response_codes:
  throttled: null
  note: >-
    No 429 was observed and none is documented; exhaustion behaviour is untested
    here. The org budget stood at usage=275 of max=6000 when probed, so
    throttling was never approached.
pagination_caps:
  note: >-
    The enforced ceiling a consumer will actually hit is the per-page record
    cap, not a request rate.
  layers:
    - layer: FHAP_Grantees
      max_records_per_page: 1000
      total_records: 889
    - layer: FHIP_Grantees
      max_records_per_page: 1000
      total_records: 1445
    - layer: RECAP
      max_records_per_page: 1000
      total_records: 5034
    - layer: RECAP_2020
      max_records_per_page: 2000
      total_records: 3559
  truncation_signal: exceededTransferLimit (boolean in the response body)
policies: []
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com