Jetstream observers disagree.
Concurrent same-region public Jetstream observers reported substantially different post volumes over the same interval.
An observed firehose volume may depend on observer choice; a study must disclose and condition on its observation source.
Result
Over the concurrent interval, jetstream1.us-east delivered 7,088 post events and each of two independent jetstream2.us-east sockets delivered 4,400. The cross-observer ratio was 1.611; the two same-observer sockets agreed at 1.000.
Design and method
On 2026-08-08, the probe opened four concurrent, post-filtered public Jetstream connections from one host for the same 120-second wall-clock interval. Counts lived only in memory and the probe wrote aggregate results; it had no event writer, database, resolver, or identity-bearing comparison set.
| observer | connections | posts observed |
|---|---|---|
| jetstream2.us-east | 2 | 4,400 each |
| jetstream1.us-east | 1 | 7,088 |
| jetstream1.us-west | 1 | 6,928 |
Limitations
This does not establish coverage, completeness, set inclusion, or which observer was closer to network truth. There was no authoritative denominator. The higher-volume stream may have been a superset, the sets may have partially overlapped, or one observer may have been temporarily degraded.
Proving set equality would require retaining per-event identity across observers. Weather Watch refuses that collection, correctly, and this finding does not weaken the boundary. Equal aggregate counts are not evidence of identical event sets.
Related instrument result: continuity
A deliberately interrupted survey retained strictly monotonic time_us across a known gap: zero timestamps decreased. Monotonic cursor time therefore did not imply complete observation. Weather Watch records gaps separately.
Related instrument result: cursor resume
Across 6/6 trials, resuming with cursor=T+1 produced exact continuation at the tested boundary. That supports the local resume mechanism; it does not refresh a historical observation or establish relay completeness.
Receipts and reproduction
The public artifacts are aggregate and identity-free:
Repository reproduction record: python3 spike/m0_probe.py control. The surrounding analysis is in M0-VERIFICATION-RESULTS.md and docs/JETSTREAM-OBSERVER-DIVERGENCE.md. Running the probe contacts public infrastructure and produces a new observation; the published receipt is not silently refreshed by later report generation.