There are two common ``cantilever'' brake designs. One lifts a wire to push the brake arms towards the rim sidewalls. The second uses long arms which are cinched together using a tension wire. The second design is often called a ``cross-wire cantilever'', ``long-arm cantilever'', or ``V-brake''.
Both styles have two common problems. First, cantilever brakes commonly squeal (makes a loud noise) when applied. Second, the brake pivot is close to the rim brake track but is offset, so moving the brake pad towards the rim also moves the brake pad towards the hub; with poor adjustment or pad wear, the pad can slip under the brake track. causing a loss of braking or the brake can jam in the spokes locking the wheel. The second problem is often called ``pad dive''.
Several Shimano brakes use a parallelogram to keep the pad aligned with the brake track. Although there is some radial motion, it is slight enough to eliminate pad dive. The extra pivots can wear substantially in a year of regular riding such that the brake rattles when the brake is off, and such that the brake pad can move several mm radially when the brake is on. The noise problem is annoying and can mask other noises which indicate other prpblems. The motion problem is not severe enough to reintroduce pad dive but can allow the pad to move radially outward so it slices the tire sidewall causing irregular braking and sudden sidewall failure.
The parallelogram design puts most load through a single bushing. Shimano sells (or at least used to sell) a set of shims (e.g, BR-M739) to install to take up as much as 0.65mm of wear. The instructions do not indicate whether multiple shim sets can be combined for additional wear compensation. That is, there is no recommended wear limit.
The other pivots in the parallelogram linkage can also wear, so increasing radial motion is still possible even after shims are installed. The directions (as of 1995) do not indicate any maximum tolerance for wear. The non-shimmed pivots are small in diameter. Although they are relatively lightly loaded it is plausible they could wear through and fail before the shimmed bushing reaches its wear limit. Failure of these other pivots would allow the pad to move radially, perhaps enough to cause pad dive or slice the tire sidwall.
An additional problem with numerous Shimano brake pads is that they gouge metal from the rim brake track much faster than other pads. The gouging is signaled by noisy braking and inspection shows metal particles embedded in the brake pad. Other brake pads can suffer the same problem but by reputation the rate of gouging with most other brands is much less.
Avid ``Single Digit'' brakes use a rigid arm. Thus, they avoid the problems of multiple pivot wear, but are subject to pad dive.
The cable clamp used in the 20 model brake (and it appears is used in other brakes) is a simple ``pinch'' clamp which crushes the cable. Unfortunately, such a clamp damages the cable. Damage occurs for several reasons. When the cable is pinched, strands are spread further apart in the plane of motion. Thus, as the arm moves, flexing is exagerated and load is carried by only a few strands. Also, the cable strands are kinked where the cable is crushed, leading to sudden stress transitions. Crushing the cable also makes it difficult to adjust the cable without causing further damage.
Although the Avid SD 20 is several years old (as of 2004), it appears the same or a similar design is still used on Avid SD brakes. A better clamp design that causes less cable damage is used on at least some Shimano brakes, for example 1998 Shimano XT.
An even better design would ``wrap'' the cable around a radius then clamp the end, so that gross motion occurs in a non-anchored portion of the cable, thus avoiding high local stresses. This sort of connection appears to be standard industrial practice. It also appears that a form-fitting clamp while theoretically inferior gives good service in practice.
The Avid SD 20 is not the only brake which uses its clamp design. It is mentioned here as a specific example, but it represents a class of brakes. Even with a better design it is still desirable to periodically check both ends of the cable for fraying that indicates pending failure.
Brakes with very high leverage typically cannot be adjusted properly to deal with pad wear. Brakes whcich have this problem include most cross-wire cantilever designs.
Consider a brake with 5mm of wearable pad material. As the pad wears, the brake must be adjusted to keep the pad surface in the same position relative to the brake lever. If the brake has 2:1 leverage, then adjusting for 5mm of pad wear requires 10mm of cable adjustment. With 3:1 leverage, adjusting for the same 5mm of pad wear requires 15mm of cable adjustment. With 4:1 leverage, 20mm is needed.
Note that the above is independent of the overall brake leverage: A brake lever with 4:1 leverage combined with a brake with 2:1 leverage gives 8:1 overall leverage. In comparison, a brake lever with 2:1 leverage and a brake with 4:1 leverage gives the same 8:1 overall leverage, but needs twice as much cable slack adjustment.
Brake designs have tended towards higher leverage. To compensate, many cantilever designs also use thinner pads. With less total pad wear (and more frequent pad replacement) the showroom brake feel is improved and the total cable adjustment can be reduced some. However, with 5mm of wear and 4:1 cable leverage (typical of post-1995 brakes) 20mm of cable adjustment is needed, but many brake levers provide only about half that. For example, Shimano XTR levers from the late 1990's have about 12mm of slack adjustment but the brake has 4:1 leverage and about 5mm of pad wear.
It seems plausible to compensate for gross pad wear by releasing the slack adjuster and moving the cable on the brake arm. However, changing the cable setting damages the cable, especially on brakes with a clamp such as the Avid SD 20 (above).
A better practice is to use two slack adjusters, for example a ``noodle'' with an adjustment range to augment the adjuster at the brake lever, or an inline adjuster.