generated: '2026-08-26'
method: searched
source: >-
  https://docs.nearspacelabs.com/error-handling, https://docs.nearspacelabs.com/production-tips
  (section "Rate Limits & Performance"), and the 4xx/5xx response sets in
  openapi/near-space-labs-tile-service.json.
description: >-
  Near Space Labs enforces a rate limit but does not publish it. The docs name 429 as the exhaustion
  status and tell clients to "back off and retry after a delay", and the page titled "Rate Limits &
  Performance" contains no limit at all — it discusses token TTL and median latency instead. No
  numeric quota, no window, no burst allowance and no rate-limit response header is documented, so an
  agent cannot budget its request rate and can only discover the ceiling by hitting it.
limit_count: 0
limits: []
exhaustion:
  status_code: 429
  title: Too Many Requests
  documented_guidance: 'Rate limit exceeded. Back off and retry after a delay.'
  source: https://docs.nearspacelabs.com/error-handling
  body: >-
    The standard error envelope {error, message, status_code, x-correlation-id} — see
    errors/near-space-labs-problem-types.yml.
response_headers:
  standard_ratelimit_headers: []
  x_ratelimit_headers: []
  retry_after: false
  note: >-
    None documented. Neither the docs nor either Swagger contract mentions RateLimit-Limit,
    RateLimit-Remaining, RateLimit-Reset, X-RateLimit-*, or Retry-After. 429 is also absent from
    every operation's declared responses in the contract — it appears only in the docs status table.
related_ceilings:
  - name: OAuth token TTL
    value: 3600 seconds (60 minutes)
    note: Not a rate limit, but the refresh cadence every client must budget for.
  - name: Static API key TTL
    value: 31536000 seconds (1 year)
  - name: Survey page size
    value: 25 by default
    note: The only published pagination ceiling.
  - name: Zoom range
    value: 14-21
    note: 'Requests outside this range return 422, not 429.'
  - name: Median tile latency
    value: ~200 ms
    note: A published performance figure, not a commitment or a limit.
provider_performance_guidance:
  - For bulk downloads, use the coverage endpoint to precompute the tile list, then parallelize fetches.
  - Use connection pooling (requests.Session() in Python) for repeated requests.
  - Batch coverage queries by AOI to find the minimal set of tiles before downloading.
gap: >-
  The provider explicitly tells consumers to parallelize bulk tile fetches while publishing no
  concurrency ceiling and no runtime rate-limit signal. That combination is the finding: a client
  following the documented bulk-download pattern has no way to stay inside a limit it cannot see.