Data snapshot: October 4, 2026, 18:13 UTC. Evidence-link totals are the public response at that read time.
A public incident record can contain 16 evidence links and still draw from one source. That is exactly what a Voidly record for Iraq dated October 4, 2026 shows: 16 linked IODA items and a source count of one. It is a connectivity disruption record. The count does not turn it into 16 independent witnesses, and it does not establish intentional censorship.
This distinction matters whenever a network failure or a suspected block is summarized in a headline. A number of links can make a claim easier to inspect, but the links need interpretation. We examined four selected records in Voidly's public Atlas incident index, spanning July 1 to October 4. This is a close reading of the published records and their link metadata. It is not an audit of every upstream measurement or a representative sample of all incidents.
Start with the question each record answers
An incident record ties a dated observation to a category, location and set of source references. That structure is useful. It gives a researcher a permalink to share and a reporter a way to state precisely which record informed a sentence. It does not automatically settle the cause, the reach across every network or the experience of every person in a country.
Voidly's index distinguishes censorship signals, connectivity disruptions and mixed cases. It also separates suspected entries from the smaller set it classifies as citable censorship or mixed incidents. These are the observatory's published classifications. Their value increases when a reader can inspect the evidence links and see how many independent sources the record counts.
The four cases below were chosen because they expose different relationships between links and sources. Their evidence counts are descriptive; they are not a league table of severity.
Sixteen links can still mean one source
The Iran record dated July 1 lists Signal, Telegram, WhatsApp and Facebook Messenger among the services at issue. It connects to 16 OONI evidence entries. Voidly classifies the record as censorship, but its corroboration field counts one source and labels the corroboration level low. Sixteen links make the source material more accessible. They do not create a second independent measurement organization.
A Russia record from the same date follows the same pattern: 16 linked OONI entries, one counted source, low corroboration. The listed services include Signal, Telegram, Tor and WhatsApp. Neither record should be shortened to “four apps were blocked nationwide” on this evidence alone. The records identify what Voidly associated with the date and where a reader can inspect the linked OONI material.
OONI itself warns that an anomalous measurement can be a false positive. Interpretation depends on the test, repeated results, the network involved and possible non-censorship explanations. A service appearing in an incident row is a lead for that examination, not an estimate of how many people lost access.
Two sources strengthen a trail without proving a cause
The Venezuela record dated September 30 brings together 41 linked items: 16 IODA references and 25 Censored Planet references. Voidly labels it a mixed connectivity and censorship incident and counts two sources. That is a different evidence shape from the July records. A reader can see two sets of observations in one place and follow the URLs attached to each item.
The extra source matters, but it does not make every conclusion available. Connectivity trouble and censorship signals can appear in the same period for different reasons. The record does not independently establish a common cause, an actor or countrywide impact.
Many of these 41 entries repeat source dashboards or portals. References are not independent measurements, and source capabilities do not verify each reference. This record alone does not establish a shared cause, an actor or countrywide impact.
The careful sentence is specific: Voidly's dated record links two source families and identifies 41 evidence items. A stronger sentence about why the event happened would require additional investigation. That boundary is a feature of good citation, because it shows where the published record ends.
Sixteen links are not sixteen outages
The Iraq record dated October 4 offers a useful counterexample to a censorship story. Its 16 linked IODA evidence items comprise 14 BGP and two ping references, across 14 region codes in the public response. Voidly classifies the entry as a disruption, and its corroboration count remains one source. The link total is a count of items from IODA. It is not a count of independent organizations, distinct outages or affected people. At the October 4, 18:13 UTC read, a separate summary said “15 critical alerts,” while the linked evidence response contained 16 items. This dated mismatch needs reconciliation; the summary is not our link count.
A practical way to cite any of these cases is to carry four pieces of information together: the record's date and category, the number of linked evidence items, the number of counted sources, and the limit on what those items establish. If the claim concerns a specific app or network, open the underlying measurement. If it concerns intent, reach or human impact, seek evidence beyond the incident row.
A public evidence trail is most useful when it makes a reader's next question easier to ask. Link totals tell us how much material a record points to. Source counts tell us something different about corroboration. The difference is where careful reporting begins.