The Pinner I test rode arrived as a used bike, and it took weeks of beating from me, as expected. On one ride the headset had slowly worked loose, but being in Finale Ligure with a café waiting at the bottom of the long trail, I continued riding and tightened it up after hydrating. The headset came loose again a few rides later, so I stopped to check and the external lower headset cup was rocking in the headtube. Upon closer inspection and removal of the fork, I discovered the headtube was slightly flared and the cup was now loose in the frame. After re-building the bike, I rode it for another few weeks sans issue – here is BTR's feedback.
We're very confident that the headset being out of adjustment is what allowed the head tube to ovalize. The loose headset allowed the lower cup to skew in the head tube - this is possible because it is an external lower cup (the bearing is a small distance away from the bottom face of the head tube), and because the bottom face of the head tube is a smaller diameter than the outside diameter of the lower bearing. This skewing (cup lifting out at the rear) caused a stress concentration (point load, instead of distributed load) at the bottom front edge of the head tube, initiating the failure. Once the head tube had started to flare, a point load at the front is inevitable, and further damage happens quickly. You might think that the interference fit of the headset cup into the head tube would prevent skewing, and it generally does, but whenever a radial load is applied to the headset, that load is essentially subtracting from the gripping load which the interference fit exerts. This effect, in combination with a loose headset, was the recipe for disaster here.
Another way to think of it is to consider the interaction between the headset bearing and the headset cup; if the headset is preloaded properly, radial load at the bearing (from leverage on the fork) will also produce an axial load forcing the headset cup against the end of the head tube, due to the angular bearing seat. If the headset isn't preloaded properly, radial load applied to the headset bearing will only result in radial load on the cup, subsequently relying on only the interference fit of the cup to the head tube to keep the cup properly in place.
If this were a customer's bike we would repair the existing head tube rather than replace it: We'll lay brazing filler inside the head tube to bring it back down below the correct size so that it can be reamed to size again. We'll also silver-braze a reinforcement ring around the outside of the head tube to further strengthen the bottom of the tube. The addition of the reinforcement ring will make the head tube stronger than it was originally, so the failure is very unlikely to occur again. To give some perspective, if it were my own personal bike, I wouldn't bother with the reinforcement ring at all. Of course, the frame will need to be recoated after that work too, so the bike will see a complete strip down and service at the same time. We'd choose to repair the existing head tube because it's a far less invasive procedure than replacing the tube; we'd also have to cut out and replace the head tube gusset if we were to replace the head tube. If the head tube was cracked or similar, we might be forced to go deeper, but in this case, we can safely just resize and reinforce the existing tube.
This is the first time we've seen a failure in this type of head tube on any of our bikes, so we know it's an isolated issue caused by the bike being ridden hard with the headset loose.