The Walk That Breaks Most Routers
You're on a video call, walking from the kitchen to the back bedroom. Signal bars look fine. Then, about halfway down the hallway, the call stutters, freezes, recovers, and you spend the next ten seconds wondering if anyone noticed.
They didn't. But you felt it.
That micro-outage wasn't a dead zone. It was a handoff that went badly, and the fact that it happens at all inside a mesh system, which is specifically sold on the promise of effortless roaming, tells you something the marketing conveniently skips.
The Node Isn't Watching You, Your Device Is
This surprises almost everyone: in a standard Wi-Fi mesh setup, the nodes don't continuously track your location and push you toward a better one. Your phone or laptop makes the roaming decision. The mesh can only nudge, suggest, or occasionally force the move.
This is baked into the 802.11 standard. Your device associates with an access point and, by default, stays there until the signal degrades to the point of desperation. That threshold, called the RSSI (Received Signal Strength Indicator) floor, varies by device and driver. A cheap Android might cling to a distant node until RSSI drops to -80 dBm or worse. A newer iPhone tends to let go earlier, around -70 dBm, which is why iPhones on a mesh network often roam more gracefully.
The practical upshot: two people with the exact same Eero or Orbi system will have completely different roaming experiences based purely on the devices they carry. Take Sofia, using a recent MacBook, gliding through her apartment without a hiccup. Marcus, on a budget Android tablet, hits the same hallway and his YouTube stream buffers for three seconds. Same nodes. Same placement. Different clients.
The hardware on the wall is almost beside the point.
What the Mesh Is Actually Doing Behind the Scenes
Mesh systems aren't passive about this. They've developed a set of tools to overcome what the industry calls the sticky-client problem: a device that refuses to roam even when a better node is right there, waiting.
The main lever is BSS Transition Management, part of the 802.11v amendment. The current node sends your device a polite but firm message: signal here is weakening, here's a list of better candidates. Most modern devices honour this. Think of it less like a handoff and more like a bouncer pointing you toward a less crowded room. Nobody gets thrown out. It's a negotiation.
Then there's 802.11k, which lets your device request a neighbour report: a list of nearby access points on the same network, their channels, and rough signal estimates. Your device can pre-survey the options before it needs to move, so when the handoff moment comes, it already knows where it's going. Latency during the roam drops significantly, often below 50 milliseconds, versus several hundred milliseconds on older systems where the device had to scan cold.
The third piece is 802.11r, Fast BSS Transition. Normal re-association involves a four-way authentication handshake that takes time. 802.11r caches security keys across the mesh so that handshake is pre-negotiated. The actual switch happens in under 20 milliseconds on a well-configured system. For voice calls or gaming, that's the difference between a glitch and silence.
All three working together, sometimes called the 802.11kvr trifecta, is what separates a real mesh system from a pair of cheap extenders sharing a network name. That distinction matters more than most buyers realise when they're comparing spec sheets at the router aisle.
The Decision Logic: What Makes a Node Win the Bid
When a handoff is actually triggered, how does the system pick the next node?
Every mesh vendor runs a proprietary algorithm on top of the 802.11 standards, which is why this part never appears on the box. But the inputs are knowable.
RSSI is the obvious one. Raw signal strength, though, is a bad sole judge. A node with a strong signal that's already handling fifteen devices and streaming a 4K feed to a TV will perform worse for your new connection than a quieter node with a slightly weaker signal. So good systems factor in channel utilization, the percentage of airtime a node is already consuming. Above roughly 70% utilization, even a strong-signal node becomes a liability.
Band steering adds another layer. Most mesh nodes broadcast on both 2.4 GHz and 5 GHz (and increasingly 6 GHz). The 5 GHz band offers faster throughput but shorter range; 2.4 GHz travels further through walls but congests easily. When your device is close to a node, the system steers it onto 5 GHz. As you move away and 5 GHz signal weakens, the system may first keep you on the same node but drop you to 2.4 GHz before eventually handing you to a different node entirely. You might roam twice in a single walk down the hall without knowing it.
Some systems, notably Google's Nest Wifi Pro with its 6 GHz backhaul, separate the client traffic bands from the inter-node communication band entirely. The nodes talk to each other on a clean 6 GHz channel while your phone operates on 5 GHz, eliminating the backhaul congestion that used to hobble earlier tri-band mesh designs.
What People Consistently Misread About Mesh Performance
More nodes always means better roaming. This is the dominant misconception, and it's wrong in a way that's almost elegant.
Pack too many nodes into a small space and your device can see four of them at nearly identical signal strength. The handoff algorithm struggles to pick a winner. The device starts oscillating between two nodes every few seconds, a failure mode called ping-ponging, and each oscillation is a brief disconnection. Your call stutters not because the signal is weak but because the system can't commit. It's like being equidistant between two coffee shops and standing on the pavement refreshing both menus.
The fix isn't necessarily fewer nodes. It's placement. Nodes should be positioned so that coverage zones overlap by roughly 15 to 20%, not 80%. A two-node system covering a 1,500-square-foot home with nodes at opposite ends will usually roam more cleanly than a four-node system crammed into the same space.
The second misconception is that mesh networks are self-optimising, so placement doesn't matter. Every vendor says this. It's partially true in the sense that nodes negotiate channels and power levels automatically. It's completely false in the sense that physics doesn't care about your vendor's app. A node placed inside a cabinet next to a microwave will degrade. No algorithm saves it.
The Wired Backhaul Advantage Nobody Talks About Enough
One specific configuration makes nearly every roaming decision faster and cleaner: wired backhaul.
When nodes communicate wirelessly, the backhaul link competes with client traffic for airtime. A handoff requires the nodes to coordinate, and that coordination takes time over a wireless link. Run an Ethernet cable between your nodes and that coordination happens at gigabit speeds with zero airtime cost. The handoff decision gets made faster, the authentication cache gets shared more reliably, and the whole 802.11kvr sequence completes before your device has time to notice a gap.
It's unglamorous. It requires drilling or running cables. Virtually every mesh vendor mentions it quietly in the fine print. Still, the performance difference in a roaming scenario can be substantial, particularly in homes with concrete walls or metal framing that attenuates the wireless backhaul signal.
If you've placed your nodes well, enabled the kvr trifecta, and you're still getting stutters on a video call mid-walk: check whether your nodes are on wireless backhaul. That's frequently the culprit.
The Part Your Devices Control That You Can't Change
There's an honest limit to all of this. You cannot force a sticky client to roam. You can configure the network perfectly, place nodes ideally, run Ethernet between every node, and a device with an aggressive roaming threshold will still cling to a dying signal longer than it should.
Curious where your device sits right now? Check your Wi-Fi signal reading while standing next to a node. At -50 dBm or better, you're winning. At -70 dBm you're in roaming territory. Below -75 dBm, you should already be on a different node, and if you're not, your device is the problem, not your mesh.
Some devices let you adjust roaming aggressiveness in their network adapter settings. Windows laptops, in particular. Mobile devices generally don't. The mesh can nudge. It cannot compel.
The real story of mesh networking isn't coverage. Coverage is the easy part, the thing you put on the box. The hard part is the negotiation: a system trying to hand off a device that doesn't entirely want to be handed off, using standards that were bolted onto a protocol designed for stationary computers. That it works as smoothly as it usually does is a small engineering triumph. The fact that a budget Android tablet can quietly undermine the whole thing is a detail that deserves more attention than it gets.