When the Tunnel Breaks: What We Learned Testing VPN Kill Switches Across 1,200 Miles of Tax Season Travel
Photo by Photo by Vitaly Gariev on Unsplash on Unsplash
A VPN is only as trustworthy as its worst moment. For most consumers, that worst moment is an inconvenience. For a CPA carrying working papers from a multi-state audit, it can be a confidentiality breach under Circular 230 and, in some states, a reportable incident.
That premise drove our latest field test. Rather than benchmarking kill switches on a lab bench, we simulated the travel conditions finance professionals actually face: unreliable hotel Wi-Fi, congested airport networks, and cellular hotspots at client sites. Across three weeks and roughly 1,200 miles, we logged every dropped tunnel and measured how each VPN responded.
Why Kill Switches Matter More Than Encryption
Encryption gets the headlines, but the kill switch is what stands between a dropped connection and exposed data. When a VPN tunnel fails, the operating system can silently fall back to the open network. Without a functioning kill switch, that fallback transmits in the clear — including any application that was mid-session.
Our methodology was straightforward. On each network, we established an active transfer of a dummy client file (synthetic data, no real PII), then forced a connection drop using three methods: network degradation via a Wi-Fi jamming tool in a controlled environment, a manual router power cycle, and a simulated carrier handoff on mobile. We measured the time between tunnel failure and full traffic blockage, and whether any packets escaped unencrypted.
Scenario One: The Hotel Network in St. Louis
Hotel Wi-Fi remains the worst-case environment. Captive portals, aggressive session timeouts, and shared bandwidth produce frequent tunnel drops — our test network at a mid-range chain property near Lambert International dropped connections an average of once every 11 minutes.
The top performers here were predictable. NordVPN and Mullvad both cut all traffic within our measured window of under two seconds using system-level firewall rules, and neither leaked a single packet across 40+ forced drops. Proton VPN performed comparably on desktop but required enabling its permanent kill switch manually — the default configuration allowed brief fallback traffic.
The disappointment was ExpressVPN, long a favorite among business travelers. Its desktop kill switch performed well, but on our Android test device, two of fifteen forced drops produced unencrypted fallback traffic lasting three to five seconds. Not catastrophic, but enough to transmit an email client's queued messages on a hostile network.
Scenario Two: The Denver Layover
Airport networks test a different weakness: roaming between access points. As your laptop hops from gate to lounge, tunnels re-establish constantly, and every re-establishment is a potential leak window.
Here, Mullvad's wireguard-based implementation was the standout, re-establishing tunnels without ever exposing the underlying interface. Surfshark also performed cleanly, though its Windows app displayed a troubling pattern of silently disabling the kill switch after system sleep — a behavior users should verify on their own machines. PIA (Private Internet Access) was solid throughout, consistent with its strong open-source track record.
Scenario Three: The Client-Site Hotspot
Mobile hotspots at client offices introduced carrier-grade NAT and signal handoffs into the mix. Most VPNs handled this adequately, but we again saw the Android gap: several providers' mobile kill switches are simply less mature than their desktop counterparts. If your firm's staff reviews returns from phones or tablets, this is the scenario that deserves your attention.
What the Timing Data Told Us
Across all scenarios, the distinction that mattered was architectural. VPNs that implement kill switches at the firewall or network-stack level — Mullvad, NordVPN, Proton VPN — blocked traffic regardless of application behavior. VPNs that rely on application-level detection occasionally lost the race, allowing a second or two of fallback traffic. Seconds matter less when you're streaming; they matter considerably when an unencrypted Outlook session is holding a client's SSN.
Recommendations for Finance Professionals
First, verify your own setup. Many kill switches are opt-in, and some revert to defaults after updates. Second, treat mobile as a separate evaluation — test your VPN on the devices your staff actually uses on the road. Third, pair the kill switch with full-tunnel mode; split tunneling, however convenient, multiplies leak surfaces.
Finally, document your remote-access configuration. As we argued in our coverage of peer review expectations, the era of 'we just use a VPN' as an adequate control narrative is closing. A tested kill switch, with evidence, is the new baseline.
