204
UTC dates with source records
Dates with retained records, not continuous uptime.
Research record / 04 September 2026
Raven System
Activity Update
Real measurements across a multi-host,
multi-domain environment
A scoped public record of what Raven measured and processed in its own environment — not a customer-performance claim.
Explore the evidence Public aggregate release
01 / Executive evidence
Measured scope.
Explicit boundaries.
≥11
Confirmed collector families
Conservative lower bound in the examined scope.
10
Measured domains
Domain presence, not uniform coverage.
501
Comparable LTE/fiber cycles
One bounded comparison population.
14 Feb 2026
Earliest producer-bound measurement
13:24:31 UTC · within the examined scope.
04 Sep 2026
Data cut
12:14 UTC · frozen audit cutoff.
Audit window: 1 February 2026 through 4 September 2026, 12:14 UTC. Retained records do not establish uninterrupted operation or equal coverage across the environment.
02 / Collection model
From measurement
to bounded evidence.
Measurements were recovered across a multi-host Raven-owned environment. Two hosts were directly audited and retained archives from a third host were assessed separately.
- Hosts
- Collectors
- Domains
- LTE / fiber
- Scoped Runtime processing
- Evidence boundary
A conceptual evidence flow, not a representation of internal deployment or a verified end-to-end decision pipeline.
Ten measured domains
- DNS
- HTTP
- RTT
- TCP connect
- TLS timing
- host health
- interface state
- jitter
- packet loss
- route/path
Retained-data scope
351 selected files · 229 byte-distinct contents · 4.86 GB of physical retained data, with copies included.
A conservative lower bound of 1,949,874 distinct payloads applies to one dataset. Neither file counts nor storage volume establish a globally unique event total.
03 / LTE and fiber
A bounded comparison.
Not a universal benchmark.
Two anonymized same-destination comparison groups from recorded observations on 8–10 June 2026. Both use the same 501-cycle population.
Comparison 1
| Access | N | Median | P95 | Max |
|---|---|---|---|---|
| Fiber | 500 | 10.980 | 11.099 | 18.046 |
| LTE | 501 | 33.447 | 67.394 | 923.652 |
+22.407 ms median paired LTE − fiber difference · 500 valid pairs
Comparison 2
| Access | N | Median | P95 | Max |
|---|---|---|---|---|
| Fiber | 501 | 10.405 | 14.124 | 14.381 |
| LTE | 501 | 30.865 | 77.114 | 1015.363 |
+20.379 ms median paired LTE − fiber difference · 501 valid pairs
Both comparison groups belong to the same 501 acquisition cycles recorded on 8–10 June 2026. They are not additive. N counts valid capture-level mean RTT observations, not individual packets or independent incidents. Historical clock uncertainty limits simultaneity claims. These bounded observations do not establish continuous uptime or general network performance.
P95 is the 95th percentile. The paired difference is not the difference of the two displayed medians. A missing valid RTT is excluded, not set to zero. Display values are rounded; the public JSON preserves audited precision.
04 / Runtime activity
Processing observed.
Outcomes not inferred.
Source validation & envelopes
624 → 208
Validated source references associated with generated envelopes in one bounded historical population.
Normalization
2,000 → 2,000
Retained inputs and normalized outputs in a separate historical processing population.
Health evaluation
24,482 and 121,522
Retained snapshots in two separate health-evaluation populations.
These populations are not summed and do not form an end-to-end funnel. Counts are not unique incidents, successful decisions or production-readiness validation.
05 / Temporal analysis
Chronology is not
decision lead.
Reconstructed threshold chronology was observed, but no qualifying evidence-ready, recommendation or approved-decision lead was established.
Retrospective threshold chronology was calculated using a frozen, past-only method. Exact detector parameters remain in the private evidence package.
Prior lead-time figures were re-audited and excluded where they were synthetic, non-reproducible, cross-target, cross-path or methodologically invalid.
06 / Evidence boundary
What the record supports.
What this proves
- Real retained measurements
- Multiple hosts in a Raven-owned environment
- Multiple measured domains
- A bounded LTE/fiber comparison
- Scoped Runtime processing
- Auditable separation of signal and decision claims
What this does not prove
- Customer validation
- ORLEN or car-wash performance
- Production readiness
- Recommendation lead
- Approved-decision lead
- Downtime reduction
- ROI
07 / Public evidence package
Read the scope.
Verify the published bytes.
Five concise content artifacts and one integrity manifest. Detailed evidence remains private. All underlying audited calculations are unchanged in this public revision.
Public methodologyAggregate metrics · JSONMeasured domains · TSVPublic claims · TSVRelease changelogIntegrity manifest · SHA-256
The manifest hashes the five content artifacts, not itself. All six files are independently checked during deployment verification. Public file hashes verify bytes, not claims.