Dunno - I would have to look at lot of diagrams to remember which bike had what kind of stem. But certainly some Cervelos, Specialized and Trek had headset cable routing that went to housing stops in the upper down tube and needed the fork removed to run new cables. Some you could get away with snagging them around the BB if they didn't cross each other and push them through the chainstay.

I'll certainly take your word that some have been done this way. In which case, seems like poor design.

It was, but full length shift housings are also considered poor design because of all the contact friction from the cable. Only department store bikes have used full length shift housings until recently.

The experimental aircraft guys recommend that in order to minimize the risk of cable failure, control cables use pulleys with at least 50x the diameter of the cable. That's not going to fit inside an integrated brake lever/shifter mechanism, so cable failure in those things is entirely expected. At least, failure of a shift cable is unlikely to be as catastrophic as failure of a control cable in an aircraft.

Yet somehow all the MTB and non-Shimano road shifters get along just fine.

They even fail on external-drum, downtube-mounted shift levers. But in that case, the impending failure is readily visible to the rider (and individual failed strands alert the rider by poking into their fingers when shifting gears). Drum diameter is inversely proportional to cable failure; the larger the drum, the less likely it is to precipitate failure. Perhaps those MTB and non-Shimano shifters use larger drums than Shimano?

I have seen a tremendous number of shifters, and cable failure at the drum is not something I would consider a feature. I would not say the road STI drum is as small as most friction levers - but they do get used more often. But so do MTB STI.

I think there is something more specific to the Shimano design contributing. And it might be the low tension on the cable because Shimano uses soft derailleur return spings, so the cables are breaking like low tension spokes do.

I'll add to John's good comments WRT frame mounted shift levers and their far less amount of fatigue breakages of the cable. Besides the cable being in easy sight all the time so any fraying can be seen ASAP, before the indexing craze these cables were both thicker and generally not stainless steel. Both aspects will contribute to a more reliable and longer lasting cable. When Shimano introduced their version of indexing (and indexing was around long before the 1980s) shifting they initially speced a thicker cable (first year DA does not use a 1.2mm cable) but quickly moved onto the current diameter for reducing friction reasons.

I consider this a great example of where advancements in features can come with a reduction of lifespan. Not a problem for those of us that are sponsored and can replace parts for free. Andy (who isn't sponsored)

I don't think I've ever had one last more than 3500 miles before breaking, and at least a couple have broken in less than 2000 miles.

I suppose that shifting frequency may be a factor.

I was looking for a number for Bowden cable, can't remember where I saw it, probably somewhere like Machinery's Handbook. 50x is higher than I recall, for some reason 38x sounds right, but with aviation it pays to add a bigger margin. I have some cables (that look like miniature aircraft cable) intended for MTB roller stems, which are high-stress and high-consequence parts. A wire rope manufacturer recommends minimum sheave diameters anywhere from 18 to 34 rope diameters, depending on the construction of the rope i.e how many wires in a strand and how many strands in the rope, those would be for cranes, elevators, excavators and other stuff that shouldn't break. Some wire ropes have a fibre core to aid flexibility, but that can also shorten the life of the rope.

Depends on how often you shift. Like Koyote, I break cables very often. I live in an area with frequent ups/downs so I believe I must shift much more often than the average bear. It’s frustrating to change a cable every 3 months (about 2500-3000 miles) but at least I have lots of practice and can swap out the old and put in a new one in about 10 minutes because practice.

I used SRAM shifting cable on Force 22 for more than 25000 km, with excessive shifting: minimum 400 times in average for a 65 km ride. I changed it although it had no signs of wear, just because many people advised changing at 5-6000 km and I was very far from that.
On the other hand, I had new Tiagra shifting cable on another bike, broken inside the shifter at less than 5000 km equivalent on a trainer, after milder use.

I wonder... do Shimano brake cables behave the same?

True for Shimano. Campagnolo has a better design and their cables typically last much longer.

Brake cables don’t, they’re not bent in the same way.