# Documented VPN Blocks, Reachable From Virginia: What Five Censorship Events Look Like From a USA Cloud

*Original FreePolicyTools research · 17 November 2026 · Two measurement sessions on Cursor-hosted USA cloud VMs*

Independent reports describe VPN protocol disruption, circumvention-website blocking, and nationwide VPN crackdowns. Those documents are about **specific national networks**. They are not a claim that the same landing pages are dark everywhere on the Internet.

We froze five documented events, then measured DNS, TCP/443, TLS, and HTTPS fetchability for twelve circumvention-tool websites and five source documents from a USA cloud VM (Cloudflare trace `loc=US`, colo `EWR`). We ran two identical cloud-direct passes (**40 probes**, collected `2026-11-17T15:07:32Z`). After Runtime Secrets injected, we repeated the same frozen URL list through ProxySeller residential **HTTP CONNECT** and **SOCKS5h** paths (**80 probes**, collected `2026-11-17T17:41:34Z`; Cloudflare egress via proxy `loc=BR`, colo `GRU`). Across conditions: **120 probes**.

**From the USA cloud vantage, every frozen circumvention host completed DNS and TLS on both passes. Ten of twelve returned HTTPS 200. The remaining two returned destination HTTP errors after a successful handshake (NordVPN HTTP 403; Outline’s `www` host HTTP 404).** The residential HTTP and SOCKS5h paths reproduced the same host-level outcome pattern: ten HTTPS 200 hosts, the same two destination HTTP errors (NordVPN title snippet “Just a moment…”, Outline “Site not found · GitHub Pages”), full documentation reachability, and expected control behavior. Pass agreement held inside every condition and transport.

CometVPN remains product and affiliate context only. It is VPN-only; we did not run a tunnel and make no tunnel claims.

These are website- and documentation-reachability findings from one uncensored cloud egress plus one residential proxy egress class. They are not an in-country censorship map, a VPN product ranking, evidence about provider logging, or a competitive uptime benchmark of ProxySeller. For background, see our [VPN versus proxy guide](/blog/vpn-vs-proxy/), [proxy routing research](/blog/proxy-routing-research/), [HTTP versus SOCKS proxies](/blog/http-vs-socks-proxies/), [how to test your VPN](/blog/how-to-test-your-vpn/), and [consumer-portal access research](/blog/consumer-portal-access-research/).

## Executive summary of measured results

| Condition | Probes | Circumvention HTTPS 200 | Destination HTTP errors after path OK | Docs | Controls | Pass agreement |
|---|---|---|---|---|---|---|
| `cloud-direct-us` | 40 | 10/12 hosts (20/24 probes) | NordVPN 403; Outline www 404 | 10/10 | OK | Yes |
| `proxyseller-residential` HTTP | 40 | 10/12 hosts (20/24 probes) | NordVPN 403; Outline www 404 | 10/10 | OK | Yes |
| `proxyseller-residential` SOCKS5h | 40 | 10/12 hosts (20/24 probes) | NordVPN 403; Outline www 404 | 10/10 | OK | Yes |
| `cometvpn-tunnel` | 0 | — | — | — | — | `NOT_RUN` |

Cloudflare location labels: cloud-direct `US`/`EWR`; residential (both transports) `BR`/`GRU`. Opaque salted egress labels differed between cloud-direct (`ip_24c91b3048e7e316`) and residential (`ip_c722c8bf3fb76b89`), confirming the proxy path was engaged without publishing raw IPs.

## What is document review versus measurement

Every claim in this article and its datasets carries an explicit layer label. Mixing layers is how readers end up believing a USA 200 OK “disproves” a Kazakhstan TLS reset. It does not. A residential 200 from a Brazil-classified egress likewise does not reproduce Myanmar DPI or Iran protocol disruption.

| Layer | Label | What it answers |
|---|---|---|
| Event catalogue | `DOCUMENT` | What OONI, Freedom House, Access Now, Filterwatch, and Human Rights Myanmar reported about named events |
| Cloud probes | `MEASURED_CLOUD_US` | Whether frozen URLs resolve, handshake, and return HTTP from this USA VM |
| Residential proxy | `MEASURED_PROXYSELLER_RESIDENTIAL` | Whether the same frozen URLs fetch over ProxySeller residential HTTP and SOCKS5h |
| Historical (pre-secrets) | `BLOCKED_PENDING_SECRETS` | Used in the first session when Runtime Secrets were absent; superseded after injection |
| CometVPN tunnel | `NOT_RUN` | VPN-only product; no HTTP/SOCKS substitute; no tunnel claim |

## The five documented events

We selected a purposive sample — recent, citeable, and centered on VPN or circumvention tooling — not a random draw of every Freedom on the Net country. Full citation fields and “what we do not claim” notes live in `events-catalogue.json`.

### E1 · Russia: fluctuating VPN protocol disruption (DOCUMENT)

OONI’s December 2024 Russia chronicle describes TSPU-assisted detection and waves of temporary, regional disruption affecting OpenVPN, IKEv2, WireGuard, and IPsec, with longer outages reported in late 2023 and continuing unevenly in 2024. Protocol blocking can leave VPN *server IP reachability* looking fine while tunnels fail — a distinction our website probes cannot settle. Our measurement layer only checks circumvention-tool *websites* as a USA/residential control. [OONI Russia report](https://ooni.org/post/2024-russia-report/)

### E2 · Kazakhstan: TLS interference and circumvention websites (DOCUMENT)

OONI’s Kazakhstan study (June 2023–June 2024 data) reports TLS interference, including MITM certificates on multiple networks, and confirmed blocking of **at least 73 circumvention-tool websites**, naming brands such as NordVPN, ExpressVPN, ProtonVPN, TunnelBear, and Surfshark. The same report notes that Tor and Psiphon *tools* often remained reachable even when related *websites* were blocked. That divergence is exactly why we refuse to treat website HTTPS 200 as tunnel success. [OONI Kazakhstan PDF](https://ooni.org/documents/OONI-Report-Kazakhstan-2024-English.pdf)

### E3 · Iran: targeted protocol crackdowns (DOCUMENT)

Filterwatch’s 2024 network monitors describe deliberate disruption of protocols used by circumvention tools, including OpenVPN-related reports, alongside policy pressure on VPN use. Those are in-path protocol findings, not website uptime charts. [Filterwatch, April 2024](https://filter.watch/english/2024/04/02/network-monitor-feb-2024-targeted-vpn-protocol/) · [Filterwatch September 2024 follow-up context](https://filter.watch/)

### E4 · Myanmar: nationwide VPN restrictions from May 2024 (DOCUMENT)

Access Now, Freedom House, and Human Rights Myanmar document a sharp escalation beginning about **30 May 2024**, with DPI-backed VPN restrictions, social-platform blocks, and large audience losses for pages that previously depended on VPN access to Facebook. Website reachability from Virginia does not speak to DPI classification of WireGuard or commercial VPN apps on Myanmar ISPs. [Access Now](https://www.accessnow.org/myanmar-vpn-ban/), [Freedom House FoTN 2024](https://freedomhouse.org/country/myanmar/freedom-net/2024), [Human Rights Myanmar — Great Firewall of Myanmar](https://humanrightsmyanmar.org/)

### E5 · Uganda 2021 election blocks (DOCUMENT, historical)

OONI’s 2021 Uganda election post records ordered blocking of social platforms and roughly **100 VPN services**, with website anomalies (including `connection_reset` patterns) for several VPN domains, while Tor/Psiphon tool tests often still succeeded on measured networks. Freedom House later notes improved VPN accessibility in subsequent coverage periods — a reminder that event timing matters and that a 2026 USA control is not a 2021 Kampala vantage. [OONI Uganda 2021](https://ooni.org/post/2021-uganda-general-election-blocks-and-outage/)

## Detailed methodology and reproducibility

### Vantage and duration

- **Platform:** Cursor-hosted cloud agent VMs (USA datacenter egress). Not Kenya PC. Cap ≤7 days; this study used two same-day sessions.
- **Cloud-direct session:** started `2026-11-17T15:07:32Z`; Cloudflare `loc=US`, colo `EWR`; opaque egress `ip_24c91b3048e7e316`.
- **Residential session:** started `2026-11-17T17:41:34Z`; secrets present; Cloudflare via proxy `loc=BR`, colo `GRU`; opaque egress `ip_c722c8bf3fb76b89` (identical label on HTTP and SOCKS5h in this session).
- **`/dev/net/tun`:** present on the host; unused. CometVPN not installed.

### Frozen targets (order fixed for all passes)

**Circumvention websites (12):** `https://nordvpn.com/`, `https://www.expressvpn.com/`, `https://protonvpn.com/`, `https://surfshark.com/`, `https://mullvad.net/`, `https://psiphon.ca/`, `https://www.torproject.org/`, `https://www.tunnelbear.com/`, `https://www.getoutline.org/`, `https://www.privateinternetaccess.com/`, `https://www.hotspotshield.com/`, `https://www.cyberghostvpn.com/`.

**Documentation URLs (5, frozen HTML fetch list):**

1. `https://ooni.org/post/2024-russia-report/`
2. `https://ooni.org/reports/circumvention/` (Kazakhstan primary DOCUMENT citation is the OONI PDF in `events-catalogue.json`; the HTML Kazakhstan slug was not used as the frozen second URL)
3. `https://filter.watch/english/2024/04/02/network-monitor-feb-2024-targeted-vpn-protocol/`
4. `https://www.accessnow.org/myanmar-vpn-ban/`
5. `https://ooni.org/post/2021-uganda-general-election-blocks-and-outage/`

**Controls (3):** `https://example.com/` and `https://www.wikipedia.org/` (expect success); `https://nxdomain-fpt-vpn-censorship-study.invalid/` (expect DNS failure).

### Cloud-direct collector (`scripts/collect_access.py`)

Per URL: DNS A/AAAA via `dig` (+time=5, +tries=1); TCP/443 connect (8s); TLS handshake with certificate subject CN/O and issuer O truncated; HTTPS GET via Python `urllib` (follows redirects), User-Agent `FreePolicyToolsResearch/1.0 (+https://freepolicytools.com/; vpn-censorship-event-study)`, first 64 KiB body SHA-256, truncated `<title>`, elapsed ms. Timeouts: DNS 5s, TCP 8s, TLS 8s, HTTP 15s. Two passes (`pass-1`, `pass-2`), same order, short inter-URL sleep.

### Residential collector (`scripts/collect_residential.py`)

Runtime Secrets (names only): `PROXYSELLER_HOST`, `PROXYSELLER_HTTP_PORT`, `PROXYSELLER_SOCKS_PORT`, `PROXYSELLER_USER`, `PROXYSELLER_PASS`. Values never printed, never committed, never placed on argv. Authentication uses a private curl `--config` file under gitignored `private/` (`proxy-user = …`). Transports: `http://HOST:HTTP_PORT` and `socks5h://HOST:SOCKS_PORT`. Flags: `--location` (match urllib redirect following), `--max-redirs 5`, connect 12s, max-time 25s, same User-Agent. Local dig/TCP/TLS are **not** treated as the residential path; cloud-direct owns those. Two passes × two transports.

![Redacted Runtime Secrets presence check during residential collection](evidence/images/proxyseller-secrets-presence-redacted.png)

*Presence-only snapshot of the five `PROXYSELLER_*` Runtime Secrets. Values are never shown.*

### What we deliberately did not do

- No CAPTCHA solving, account creation, VPN installer downloads, or third-party tunnel establishment.
- No User-Agent rotation mid-study to bypass WAF/bot challenges.
- No amendment of the frozen Outline `www` URL into the main matrix after observing apex-vs-www differences.
- No invented credentials when secrets were previously absent.
- No competitive benchmarking of ProxySeller uptime or anonymity.
- No CometVPN as HTTP/SOCKS.

### Reproduce

```bash
python3 scripts/collect_access.py          # cloud-direct
python3 scripts/collect_residential.py     # requires PROXYSELLER_* secrets
python3 scripts/analyze.py                 # public tables + mirror + ZIP
```

## Finding 1: USA cloud reachability is the wrong test for in-country VPN blocks

If a report says Kazakhstan interfered with TLS to `nordvpn.com` on local networks, a successful TLS handshake from Virginia does not refute that report. It confirms the more boring fact that the destination still speaks TLS on the public Internet.

Our two cloud-direct passes agreed for every primary host. DNS resolved; TCP/443 connected; TLS completed for all twelve. **Eleven of twelve hosts negotiated `TLSv1.3` on both passes; `psiphon.ca` negotiated `TLSv1.2` (cipher `ECDHE-ECDSA-CHACHA20-POLY1305`) on both passes.** That pattern is what an uncensored control vantage should look like for popular HTTPS sites. It is useful precisely because it **fails to mimic** the anomalous classes OONI uses when confirming blocks (timeouts, connection resets during handshake, forged certificates).

**Cloud-direct elapsed times (reachable circumvention HTTPS probes only):** mean ≈ 285 ms (median ≈ 190; min 40; max 724; n=20). These are researcher-fixture timings on this VM, not a product SLA.

Readers comparing a USA speed-test screenshot to an OONI anomaly chart are comparing different networks. Keep the labels. Our [VPN testing guide](/blog/how-to-test-your-vpn/) covers tunnel checks that website fetches alone cannot replace.

## Finding 2: ten of twelve sites returned HTTPS 200; two failed at HTTP after TLS

Host-level outcomes below are from both cloud-direct passes and both residential transports (HTTP + SOCKS5h). Pass agreement held inside every condition. Residential HTTP and SOCKS5h agreed with each other and with cloud-direct on outcome class for every frozen host.

| Host | Cloud | Residential (HTTP + SOCKS5h) | HTTP | TLS (cloud) | Notes from evidence |
|---|---|---|---|---|---|
| www.expressvpn.com | reachable | reachable | 200 | TLSv1.3 · Amazon | Title “ExpressVPN: Best VPN Service…” |
| protonvpn.com | reachable | reachable | 200 | TLSv1.3 · Let’s Encrypt | Cloud title empty; residential “The best VPN for speed and security” |
| surfshark.com | reachable | reachable | 200 | TLSv1.3 · Google Trust Services | Title includes identity/scam-protection wording |
| mullvad.net | reachable | reachable | 200 | TLSv1.3 · Let’s Encrypt | Title “Mullvad VPN - Privacy is for the people” |
| psiphon.ca | reachable | reachable | 200 | **TLSv1.2** · Google Trust Services | Only frozen host on TLSv1.2; title “Psiphon” |
| www.torproject.org | reachable | reachable | 200 | TLSv1.3 · Let’s Encrypt | Title “Tor Project \| Anonymity Online” |
| www.tunnelbear.com | reachable | reachable | 200 | TLSv1.3 · Sectigo | Title “TunnelBear: Secure VPN Service” |
| www.privateinternetaccess.com | reachable | reachable | 200 | TLSv1.3 · GoDaddy | Title “Private Internet Access…” |
| www.hotspotshield.com | reachable | reachable | 200 | TLSv1.3 · Amazon | Title “Hotspot Shield…” |
| www.cyberghostvpn.com | reachable | reachable | 200 | TLSv1.3 · GoDaddy | Title empty in captured snippets |
| nordvpn.com | tls_ok_http_fail | tls_ok_http_fail | 403 | TLSv1.3 · Sectigo | Cloud body empty/challenge-style; residential title “Just a moment…” |
| www.getoutline.org | tls_ok_http_fail | tls_ok_http_fail | 404 | TLSv1.3 · Let’s Encrypt | Residential title “Site not found · GitHub Pages” |

The NordVPN and Outline cases matter for methodology. A naive “site blocked?” script that only checks HTTP status would mark them failures and invite a censorship narrative. Transport evidence says otherwise: name resolution and the TLS handshake succeeded. The HTTP layer then answered with an application status — closer to destination policy, bot challenge, or URL shape than to a national TSPU reset.

We did not bypass the 403, change User-Agent mid-study, or amend the frozen Outline URL into the main matrix. A separate apex check to `https://getoutline.org/` returned HTTP 200 while `https://www.getoutline.org/` remained 404; that check is interpretive context only (`evidence/outline-apex-vs-www-interpretive.json`) and is **not** in the 120-probe totals. See also [how to unblock websites with VPNs or proxies](/blog/how-to-unblock-websites-with-vpns-or-proxies/).

## Finding 3: the documentation of censorship remained fetchable

All five documentation URLs returned HTTPS 200 on both cloud-direct passes and on both residential transports (10/10 probes per transport; 30/30 documentation probes across conditions).

| Documentation URL | Cloud HTTP | Residential HTTP / SOCKS5h | Captured title (truncated) |
|---|---|---|---|
| `https://ooni.org/post/2024-russia-report/` | 200 | 200 / 200 | Censorship Chronicles: The systematic suppression of independent media… |
| `https://ooni.org/reports/circumvention/` | 200 | 200 / 200 | Research reports: Blocking of Circumvention Tools \| OONI |
| `https://filter.watch/english/2024/04/02/network-monitor-feb-2024-targeted-vpn-protocol/` | 200 | 200 / 200 | Targeted VPN Protocol Disruption Following New Legislation |
| `https://www.accessnow.org/myanmar-vpn-ban/` | 200 | 200 / 200 | Worse than China or Iran? Myanmar’s dangerous VPN ban |
| `https://ooni.org/post/2021-uganda-general-election-blocks-and-outage/` | 200 | 200 / 200 | Uganda: Data on internet blocks and nationwide internet outage… |

That is not a human-rights victory metric. It shows that **records about blocking** and **paths inside blocked networks** are different objects. A researcher on USA cloud can read the Kazakhstan PDF (DOCUMENT citation) while being unable to reproduce Kazakhstan’s TLS MITM. The frozen HTML fetch list uses the OONI circumvention-reports index as the second documentation URL rather than a Kazakhstan HTML slug that was not stable at freeze time — the Kazakhstan PDF remains the primary DOCUMENT source in `events-catalogue.json`. Publication reachability is necessary for open science; it is not a substitute for [OONI Probe](https://ooni.org/install/) vantage points.

## Finding 4: controls confirm the collector distinguishes success from DNS failure

`example.com` and `www.wikipedia.org` completed reachable outcomes on both cloud-direct passes and on both residential transports (HTTP + SOCKS5h). The synthetic `.invalid` name produced the expected DNS failure class on cloud-direct (TCP/TLS/HTTP skipped) and expected resolve/fetch failure through the proxy on residential legs. Without that negative control, empty DNS answers and true blocks are easier to confuse.

## Finding 5: ProxySeller residential HTTP and SOCKS5h reproduced the same host-level pattern

Runtime Secrets were present for all five required names. We unmarked `BLOCKED_PENDING_SECRETS` and ran the frozen URL list through residential HTTP and SOCKS5h.

![Cloud-direct versus residential outcome classes](evidence/images/cloud-vs-residential-outcomes.png)

*Host-level HTTPS outcome classes for the twelve frozen circumvention sites across cloud-direct and both residential transports.*

**Elapsed times (residential, reachable circumvention probes only):** HTTP mean ≈ 3459 ms (median ≈ 3076; min 2005; max 6674; n=20). SOCKS5h mean ≈ 4279 ms (median ≈ 3472; min 1855; max 10265; n=20). These are researcher-fixture timings on this session, not a ProxySeller SLA claim.

Three bounded observations:

1. **Egress class changed; outcome class did not.** Cloud-direct Cloudflare traces showed `US`/`EWR`. Residential traces showed `BR`/`GRU`. Opaque egress labels differed from cloud-direct, confirming the proxy path engaged.
2. **HTTP and SOCKS5h agreed** on host-level reachability classes for this frozen list. That is a result for these destinations and this session, not a claim that HTTP and SOCKS are interchangeable for every application ([HTTP versus SOCKS guide](/blog/http-vs-socks-proxies/)).
3. **Residential success is still not in-country evidence.** A Brazil-classified residential exit fetching `nordvpn.com` with HTTP 403 after TLS says nothing about TSPU OpenVPN disruption or Myanmar DPI.

If you need that egress class for your own researcher-operated fixtures, compare [ProxySeller residential options](/proxyseller/) and verify the path yourself. Affiliate status did not alter measured statuses.

## Finding 6: CometVPN tunnel legs remain explicitly unfinished

CometVPN is VPN-only. Even though `/dev/net/tun` exists on this host, we did not install or claim a CometVPN tunnel. Articles may still link [`/cometvpn/`](/cometvpn/) for product context. Tunnel efficacy against the five DOCUMENT events remains unmeasured here. **Do not treat residential proxy results as a substitute for VPN tunnel tests.**

## Session timeline

| Time (UTC) | Action | Result |
|---|---|---|
| Earlier same day | Initial PR #2 cloud-direct package | 40 probes; residential `BLOCKED_PENDING_SECRETS` (secrets absent on that agent) |
| ~15:07 | Cloud-direct collection on replacement agent | `US`/`EWR`; 10/12 HTTPS 200 |
| Secrets injection | New agent start after Runtime Secrets attached | All five `PROXYSELLER_*` names present |
| ~17:34 | Residential smoke test (example.com + ipify) | HTTP + SOCKS5h 200; egress label ≠ cloud-direct |
| ~17:41 | Full residential matrix with `--location` | 80 probes; host-level pattern matches cloud-direct; `BR`/`GRU` |

DOCUMENT claims were never rewritten. Only condition status and measured residential evidence were amended.

## How the failure classes differ from OONI block signatures

| Observation in this study | Closest naive reading | Why that reading is wrong |
|---|---|---|
| NordVPN HTTP 403 after TLS | “NordVPN blocked in USA / via residential” | DNS + TLS succeeded; HTTP application/challenge response (`Just a moment…` on residential) |
| Outline `www` HTTP 404 after TLS | “Outline blocked” | TLS OK; GitHub Pages “Site not found” for frozen `www` host; apex separately 200 |
| USA or BR residential HTTPS 200 | “Kazakhstan / Myanmar not blocking” | Wrong network; DOCUMENT events describe other ASNs and time windows |
| Docs fetchable | “Censorship overstated” | Publication reachability ≠ in-path interference |

OONI’s public methodology and Explorer remain the correct tools for in-country confirmation. [OONI Explorer](https://explorer.ooni.org/) · [OONI measurement docs](https://ooni.org/support/)

## What this study cannot tell you

- **Not an in-country map.** No RU/KZ/IR/MM/UG ISP vantage; no OONI Probe probes from those networks in this package.
- **Not tunnel usability.** Website HTTPS ≠ WireGuard/OpenVPN/IKEv2 success under DPI.
- **Not provider ranking.** Twelve purposive marketing sites; not market share, trust, or logging claims.
- **Not ProxySeller uptime/anonymity.** One purchased residential package, one session, researcher fixtures only.
- **Not CometVPN performance.** `NOT_RUN`.
- **Not longitudinal.** Single-day sessions; sites and WAF policies change.
- **Not browser realism.** Scripted UA; no WebRTC/STUN, no CAPTCHA solve, no human browsing.
- **Not exhaustive documentation.** Five event docs + catalogue citations; not every NGO report on each country.

## Evidence inventory (public vs private)

| Public (in git / `/research/vpn-censorship-event-study/`) | Private (gitignored `private/`) |
|---|---|
| `protocol.md`, `events-catalogue.json`, `findings.md`, `research-article.md` | Raw egress IPs, curl proxy-user configs |
| `evidence/cloud-direct-*.json`, `proxyseller-residential-*.json` (redacted) | Body binaries from curl fetches |
| Host CSVs, `analysis.json`, `summary.json`, `env-discovery.json` (presence only) | Any secret material |
| `images/proxyseller-secrets-presence-redacted.png`, `images/cloud-vs-residential-outcomes.png` | — |
| `outline-apex-vs-www-interpretive.json` (not in main totals) | — |
| `public-research-package.zip` + SHA-256 | — |

[Download cloud-direct dataset](evidence/cloud-direct-latest.json) · [Download residential dataset](evidence/proxyseller-residential-latest.json) · [Analysis](evidence/analysis.json) · [Outline apex note](evidence/outline-apex-vs-www-interpretive.json) · [Event catalogue](events-catalogue.json) · [Protocol](protocol.md) · [Findings](findings.md) · [ZIP package](public-research-package.zip)

## Practical takeaways

1. **Read the vantage.** “VPN blocked” almost always means “blocked on these networks at this time,” not “offline globally.”
2. **Separate website reachability from tunnel usability.** Kazakhstan’s OONI write-up shows tool reachability and website blocking can diverge.
3. **Inspect the failure class.** HTTP 403 after TLS is not the same evidence as a handshake reset.
4. **Use open measurement corpora for in-country claims.** OONI Explorer and national partners remain the right tools for RU/KZ/IR/MM/UG path evidence.
5. **Choose tools for the job.** A [VPN](/cometvpn/) and a [residential proxy](/proxyseller/) solve different routing problems. This study validates neither against DPI; it shows how control and residential website fetches behave for a frozen public URL list.

## Affiliate disclosure

FreePolicyTools may earn commissions from [ProxySeller](/proxyseller/) and [CometVPN](/cometvpn/) links. No vendor funded, edited, or pre-approved this measurement session. Affiliate status does not change the `DOCUMENT`, `MEASURED_CLOUD_US`, `MEASURED_PROXYSELLER_RESIDENTIAL`, or `NOT_RUN` labels, and it does not authorize inventing credentials or claiming unrun tunnel results.
