The sixty-second answer
Your desk phones stop, because they need a connection to reach the system. Your number does not stop, because it lives in the network rather than in your building. Callers keep getting through only if a failover rule already exists sending calls to mobiles, another site, or voicemail.
Two different things stop working
An outage feels like one event and is actually two, with different consequences and different fixes.
Your handsets go quiet. They light up, they show the time, they do nothing else. Nobody in the building can make or take a call on them.
Your callers still call. They have no idea anything happened. Your number is still advertised, still listed, still on the van. Calls keep arriving at it all day.
The whole question is what happens to those calls in the gap. That is a routing decision, and the uncomfortable truth is that it is made in advance or not at all. Nobody configures failover during an outage, because configuring it usually requires the thing that is broken.
Why your number survives and your phones do not
A phone number on a hosted system is an address held in the network. It is not a wire into your office, and there is nothing in your building for the outage to take away from it.
Your handsets are a different matter, because they work by announcing themselves. A device sends a registration message when it starts up and at regular intervals afterwards, associating the user with the machine they are currently logged into, and that binding gets written into the directory the system consults to route incoming calls [1]. When your connection dies, the bindings stop being refreshed. Within a short window the system knows those desks cannot be reached.
That moment is your opportunity. The system is aware the office is unreachable, and it is holding a live call for you. What it does next is whatever rule you left behind.
The rules worth leaving behind
Ring mobiles. The strongest and cheapest option. A business extension installed on the phones your people already carry reaches the system over the mobile network, so it is unaffected by your office connection. This works even better if those mobiles are already ringing alongside the desk phones day to day, because then there is nothing to switch on and nothing to remember. The mechanics are in answering business calls on your cell phone.
Send calls to another site. If you have two locations, each is the other's backup, and callers to either number reach a person. Setting this up is ordinary configuration rather than a project; see transferring calls between two locations and one dial plan across sites and remote workers.
Answer with something honest. Failing everything else, callers should reach a greeting and a mailbox rather than ringing forever. Messages delivered as email can be read from a phone on mobile data while the office is dark, which is the practical version of staying open; see voicemail to email for Canadian businesses.
Have a way to change the greeting from outside. The single most useful thing during a long outage is being able to record a new message from a mobile: what happened, when you expect to be back, what to do meanwhile. Check you can do that before you need to. Related patterns are in after-hours and holiday call routing.
Outbound calls are a separate problem
Everyone plans for incoming calls and forgets that the business also needs to ring people back. During an outage that happens from mobiles, and the thing to check in advance is which number the customer sees. If your team ends up ringing customers from personal mobiles, those numbers are in your customers' phones permanently, and the next call about that job comes to somebody's pocket instead of your business. A business extension on a mobile presents the business number, which is the entire reason to have one there.
The failure nobody plans for: the connection that half works
A dead connection is easy. The bindings lapse, the failover fires, calls go to mobiles.
The bad case is the connection that is technically up while losing or delaying traffic. The phones still look registered, so nothing fails over, and calls are delivered to desks where they cannot be heard. This is worse than an outage because it is invisible from inside the building and looks like working service from outside it.
The reason it is so unforgiving is structural. Live voice is carried by a protocol that explicitly does not guarantee delivery, does not prevent out-of-order delivery, and does not assume the network underneath it is reliable; it numbers the pieces so the far end can put back together what actually arrived [2]. There is no second attempt for a lost fragment of a sentence. If you are chasing this symptom rather than a clean outage, start with why business calls keep dropping, and check the connection expectations in bandwidth and quality of service requirements.
Emergency calls deserve their own written answer
Do not infer how emergency calling behaves during an outage, and do not accept a verbal answer about it. Ask your provider, in writing, exactly what happens to an emergency call from each device when the office connection is down, and what address is recorded against every phone and every backup path. Then confirm those addresses are current, especially after a move, and re-check them whenever you add a device. This is the one part of an outage plan where a wrong assumption is not recoverable, and it is worth the email.
What about a backup internet connection?
A second connection from a different provider keeps the desk phones alive through a single-provider outage. It is genuinely useful for a busy front desk, a clinic, or anywhere the number of calls makes mobiles impractical.
Be clear-eyed about what it is not. It does not help in a power cut unless the equipment is also on battery. It is a monthly cost for something you use rarely. And it does nothing about the half-working connection described above, which is a quality problem rather than an availability one. For many small offices, failing over to mobiles is most of the benefit at none of the cost. If you are weighing where the equipment should live in the first place, that trade-off is in hosted versus on-premise phone systems in Canada.
Test it on a Tuesday morning
Once a quarter, unplug the office router at a quiet hour and call your own main number from outside. Time how long before the call reaches somewhere sensible. Check whether the caller hears anything odd first. Check that whoever answers can ring back with the business number showing. Then plug it back in and write down what happened.
Fifteen minutes, four times a year, and you will find the setting nobody updated when the last person left. Any provider who tells you their service keeps working through an outage is making a performance claim, and paragraph 74.01(1)(b) of the Competition Act makes such a claim reviewable where it is not based on an adequate and proper test, with the proof of that test resting on whoever made it [3]. Your own quarterly test is the version of that evidence you control.
Where we sit
We run hosted phone systems for Canadian businesses, and we set the failover rules up at install rather than selling them later: where calls go when the office cannot be reached, which mobiles ring, what the caller hears, and how a greeting gets changed from outside the building. It is configuration, not a product tier.
We will also tell you plainly what we do not control, which is your building's internet and its power. Telephone service is treated in Canadian law as infrastructure rather than a convenience, with section 7 of the Telecommunications Act affirming its essential role and setting out the objective of an orderly system throughout Canada [4]. Most of the 1.08 million small businesses in this country [5] have no continuity plan for the day their connection drops, and for a phone system the plan is four settings and a quarterly test. What is included, and how extensions are counted, is on our pricing page.