Exhibit Dossier

Detailed technical records, CDR statistics, quality-metric analysis, and support logs regarding Ticket #387285. This page carries the full evidentiary detail behind the executive summary.

Exhibit A — Key exchange

OpenFax named the cause within the hour. Telnyx support said it didn't exist — twice.

On the night of June 29, 2026 at 23:29 ICT, OpenFax forwarded Telnyx's own “Voice services update – all regions” maintenance notice back to Telnyx support and asked directly: “Is incident update related?”

“I've checked the Telnyx Status Page and found no published incident specifically named 'Voice services update - all regions.'” Telnyx support — June 29, 23:31 ICT
“I don't see a specific 'Voice services update – all regions' incident currently posted on the public status page.” Telnyx support — June 30, 23:33 ICT

Both replies listed only unrelated wireless/roaming incidents as active. This is the same maintenance window that appears in the public status-page anchors on the executive timeline (June 29 entry). OpenFax identified the correct cause within roughly an hour of the notice going out — and was told by Telnyx's own support that it did not exist.

The outage fingerprint: MOS collapse

MOS (Mean Opinion Score) is a direct, objective measure of whether audio was actually flowing. Average MOS sat rock-steady at 4.44–4.49 every single month from August 2025 through June 26, 2026 — no drift, no gradual decline. Then, hour by hour:

Time (UTC)Avg MOSInterpretation
Through Jun 264.44–4.49Normal — eleven months of flat baseline
Jun 29, 11:00 onward~1.0–1.5Audio effectively unusable — earliest bad reading in the data
All of Jun 30~1.0–1.5Still broken
Jul 1, 04:00–05:001.000Still broken (only 4 calls logged)
Jul 1, 13:00 onward4.44–4.49Snaps back to baseline instantly — exactly at the AWS/regional cutover
Aftershock, not a second failure

The abandoned-call rate stayed elevated (30–35%) for roughly four more hours after MOS recovered, only normalizing by July 2 — consistent with backlog/retry traffic clearing, not a new fault.

Diagnostic sequence — June 30

How OpenFax isolated the failure, step by step, on its own.

Support call record — Ticket #387285

Key facts established across the support engagement.

CDR trend: abandoned-call rates

Monthly abandoned-call rate against the 20% surcharge threshold, August 2025 through the July 1–6 outage window. Notably, the rate was already at 16.09% in September 2025 — a month before the October 28 policy announcement — i.e., the pattern predates the policy rather than resulting from it.

The 20% billing threshold was exceeded in Nov '25 (23.85%) and Jun '26 (20.17%). The July 1–6 window averaged 22.64%, driven almost entirely by July 1's record daily spike of 32.53% during the outage.

Invoice & policy reconciliation

Invoice-billed surcharge units, cross-validated against an independent recompute from raw CDRs.

MonthInvoice-billed abandoned callsIndependent CDR recomputeVarianceAmount billedRate
Nov 2025512,752512,7502 calls (≤0.01%)$2,563.7623.85% — over threshold
Jun 2026339,966339,93630 calls (≤0.01%)$1,699.8320.17% — over threshold
Jul 1, 2026(during outage window)32.53% daily peak
Why the two counts are shown side by side

The invoice figures and the CDR-derived figures (Hangup Details = recv_cancel) are independently sourced and agree to within 0.01%. This cross-validates the CDR analysis against what Telnyx actually billed: the fee calculation is not in dispute — the question is whether outage- and configuration-driven abandonment should be exempt from it.

Policy details

Telnyx issued the abandoned-call surcharge notice on October 28, 2025, effective November 1, 2025: $0.005 per abandoned call, applied only when the monthly abandoned rate exceeds 20%. Telnyx's own confirmed definitions: a Short Duration Call is under 6 seconds; an abandoned call is one where the originating side disconnects during ringing/setup, before answer.