In Cisco's Architecture for Voice Video and Integrated Data (AVVID), there is a feature known as Survivable Remote Site Telephony, which gives you the ability to have SRST-enabled IP phones to register with a router in the event that they lose network connectivity to the Call Manager. Generally the way this is deployed is for remote sites that are small enough that you can't justify a local Call Manager server for them, but that are still important enough to require a fault-tolerant telephony solution.
Unfortunately, SRST had a pretty major limitation, which is that the router that takes over call processing for the site had to be the default gateway for the site. This is fine in many smaller remote offices, where you might have just a few users and a single subnet, but for many organizations, remote offices are often factory floors or warehouses, which have few telephone handsets, but are very large physically, or have other wiring issues that make designers prefer routing instead of switching. Other organizations have remote offices with hundreds of users, but may still prefer to use a centralized Call Manager farm. So with the massive popularity of layer-3 switching, it's not unusual to find many "remote" offices that have users scattered around on many subnets. Further, popular designs for redundancy, and even Cisco's own "Core-Distribution-Access" layered architecture model conflict with the basic restriction of SRST that all the users at a site share the same physical default gateway.
Fortunately, the more recent versions of Call Manager do not have this requirement. As of version 3.3, you can make the decision as to which router processes calls in the event of a WAN failure on a device-pool basis, and you can specify a device that is not the default gateway.
Thomas Alexander Lancaster IV is a consultant and author with over ten years experience in the networking industry, focused on Internet infrastructure.