6 min read · Published August 24, 2026 · Vast Networks
What you are actually buying
The purpose of a second connection is that it survives whatever takes the first one down. That makes the useful question not "do we have two circuits" but "what could remove both at once". Every shared element between the two paths is a way for one event to take out both, and shared elements are easy to acquire accidentally because nothing in the ordering process reveals them.
Work through the chain from your equipment outward: the router, the building entry, the conduit or aerial path to the street, the route back to the provider’s facility, and the provider’s own upstream connectivity. Independence at every layer is rarely achievable or affordable. Independence at the layers most likely to fail — physical path and building entry — is usually both.
The failure modes worth designing against
**The backhoe.** Physical damage during construction is the most common cause of a hard outage, and it takes out everything in the trench. Two circuits in one conduit fail together, whoever sold them.
**The building entry.** Many buildings have exactly one point where services enter. Two providers, two routes across town, one wall penetration — and a flooded riser or a contractor with a core drill removes both.
**The upstream.** Two providers can buy transit from the same upstream carrier. Their outage becomes your outage twice over. Asking is the only way to find out.
**Your own equipment.** A single router or firewall with two WAN ports is a single point of failure holding two circuits. This one is frequently the cheapest to fix and the most often overlooked.
**Power.** Both circuits terminate on equipment that needs power. If the UPS covers one rack and not the other, the redundancy has a runtime limit nobody documented.
Choosing the second path
The strongest second path is a physically separate route entering the building at a different point, from a provider with independent upstream connectivity. Whether that exists at your address is a question about conduits, crossings, and entries rather than about contracts, and it is covered properly in why fiber route diversity matters — including the four ways two circuits that look separate turn out not to be.
The option that guide does not cover is changing medium. A fixed wireless link fails for entirely unrelated reasons: it has no trench, no conduit, and no duct to share. As a backup path it is frequently more genuinely independent than a second fiber that follows the same street, and its lower capacity matters less than it appears — during an outage the question is what still functions, not what performs identically.
Where neither is available, the honest position is a single path with a documented recovery plan: mobile failover for a defined subset of traffic, a work-from-home arrangement, or an accepted downtime window agreed in advance. Knowing you are on one path is materially better than assuming you are not.
Failover that actually works
A second circuit that requires someone to reconfigure a router is a recovery procedure, not failover. Decide in advance what should switch automatically, what should be manual, and how long each takes. Then decide what happens to sessions: voice calls will drop on an IP change unless the platform handles registration migration, and VPN tunnels will need to re-establish.
Test it deliberately. Pull the primary circuit during a scheduled window and observe what breaks — not once at installation, but periodically, because configurations drift and equipment gets replaced. The most common discovery is not that failover fails, but that it works while something depending on a static address quietly does not.
How much redundancy is proportionate
Cost the outage before designing the solution. A distribution operation that cannot ship, a clinic that cannot retrieve imaging, or a plant whose batch record must be complete has an hourly cost that makes diverse paths obviously worthwhile. An office where people can work from home for an afternoon does not.
Between those poles, the useful exercise is to identify which specific functions must survive and provide for those rather than for the whole site. A backup path sized for the point-of-sale, the phones, and the essential applications is far cheaper than one sized to be indistinguishable from the primary, and it covers the actual requirement.