See also also bike pic for other bike-related technical discussions and failed parts.
There is a common idea that frames will “go soft” or wear out when they are used hard for a while.
One story (perhaps apocryphal) is of a frame builder “long ago in Italy” who would receive people complaining their frame had “gone soft”. And: could he do anything about it? He would inspect the frame and sometimes find a crack, etc. Other times, he would say “I can do something, but you'll have to leave it here a couple weeks while I work on it.” They would return and be delighted with the result, but he would never disclose exactly what was the repair. Well, years go by and the master retires and finally admits that all he did was clean the bike and put it in the back. The rider, after a couple weeks off the bike (and, likely, riding some other bike), would approach the bike “fresh” and thus be as happy as when it was fresh for them years before.
In other words, frames don't go soft, people get used to them and their perceptions change.
That said, frames do wear out.
Here is one that was on a car rack for a long trip and was rubbing as the car went down the road:
(Photos courtesey of Yellow Jersey.)
The frame is aluminum — relatvely soft and not-wear-resistant compared to steel, but probably more wear-resistant than carbon fiber.
Aluminum is often hard to repair because welding weakens it. It can be re-strengthened by putting it a special oven for a heating/cooling process called “heat treating”, but that means stripping the paint and supporting the frame on a rack for the process.
The frame above is inexpensive, and the rider is not too vigorous/demanding, so “just ride it and keep an eye on it.”
"Yield" is "it just plain wasn't strong enough".
"Fatigue" is there are tiny microscopic cracks in everything, and some loads are small enough they do not bust the whole piece of metal, but are big enough to slightly grow one of the microscopic cracks. Over enough of those loads, there are enough cracks that have grown far enough that the part finally busts.
Bike frames in particular tend to have some specific high-load points. For example, around the bottom bracket. Everytime you press on the pedals, it twists the bottom bracket, and after a while the seat tube gets fatigued at the point of highest leverage — right where it enters the bottom bracket, which is also where the metal has been weakened slightly by brazing or welding. For an example, see [here as of 2015/02].
Another high-load point is the right rear dropout where the chain stay connects. Even a "stiff" axle bends slightly on every pedal stroke; that twists the dropout, and in turn tends to fatigue the dropout, Again, at the point of highest load, which is also where it has been weakened by welding. For an example, see [here as of 2015/02].
As a rule of thumb, "if you ride it enough, it will fatigue". Frames wear out.
However, there are also things frame designers and builders can do to minimize fatigue. Again, fatigue is basically repeated loads which are enough to grow microscopic cracks, so things which reduce the load at the crack level will reduce fatigue.
Some examples include:
Careful design — changing joint design can change joint load. For example, slightly ovalizing the seat tube where it joins the bottom bracket shell can improve the seat tube's leverage on the bottom bracket shell, which in turn reduces the load a crack sees on any given pedal stroke.
Unfortunately "careful design" is a tricky topic: many people have introduced changes that "look smart" yet wind up breaking faster than the original "un-improved" design. One observation here is people can design well if either they have the right background and go through a lot of careful engineering, or if we stick with time-tested designs; but that even small changes to a time-tested design sometimes leads to much faster failures.
Less heat — the metal is weakened by heating, so heating to a lower temperature means (all else equal) stronger metal and in turn it takes a higher laod to grow a crack. Brazing uses lower temperatures than welding, and some braze metals (e.g., silver) have lower braze temperature than others (e.g., brass). Training and practice counts here, too, as it is easy to buy low-temperature braze material, harder to get the metal only as hot as needed and no more.
Local heat — the metal is weakened by heating, so (relatively) quick heat-up and cool-down heats less of the metal. This sometimes leads to funny trade-offs. For example, welding has higher heat, but may have a much smaller heated area (more local). In turn, you can sometimes use a tube with a shorter but thicker butt and wind up with a welded joint as fatigue-resistant as a brazed joint and with the same total material. Again, training and practice counts here as welding technique affects the size of the heated region.
Better material — the metal is stronger, so it takes a higher load to grow a (microscopic) crack. Put another way, you pedal just as much and just as hard, but fewer of the pedal strokes rise to the level of crack growth. Note that "better material" is often accompanied with "less material" to save weight, thus putting you back more or less where you started.
More material — the metal is damaged by high load. Use twice as much metal and you should get half the load on the (microscopic) crack. Fortunately, the highest loads are mainly localized to places like the end of the tube, so you can double the metal at the end of the tube with almost no change in total weight. There are at least four common approaches to "more material":
More flex, somewhere else — here, the idea is the load at the joint depends on the stiffness of all the pieces. If the whole length of the seat tube is stiff, it leads to some load at the joint with the bottom bracket. If the seat tube is floppy, then as you push on the pedals and load goes up, the seat tube flexes and so limits the peak load on the joint, and thus the peak load on the microscopic cracks. As another example, a long dropout may flex along its length, which reduces the peak load at the joint with the chainstay.
As a practical matter, it is fairly straightforward to combine the observations above and to build frames durable enough to last decades of daily use. Further, fatigue-resistant construction also tends to be stronger (for a given weight) against one-time "yield" loads like a crash.
It does add some cost to use a sensbile design, assemble a frame with butted tubing and/or lugs/gussets/fillets, and do it all with a skilled builder.
However, at almost any price, there are examples of trade-offs. There is often little extra cost to using a good design, and adding just a few gussets in the right place may lead to a big improvement in durability. Conversely, there are examples of "department store" bikes sold at low cost with designs that seem aimed at the twin goals of (a) look flashy to sell one today, and (b) break as quickly as possible to get the owner back next year to buy another.
Unfortunately, it is not assurance of durability to move to more-expensive bicycles, whether mass-market manufacturers or established custom builders. It may take a few years of regular use before a "novel" design starts to show durability problems, and if the maker feels every frame must include "this year's newest trend", there is bound to be an oops here and there.
A reputation for good warranty support can help, but some new makers may have good products (with no history) while existing makers may be about to go out of business (and have quietly changed their tried-and-true designs to save cost and not go out of business).