First, let me lead off by saying that I don't have any inside information.
Second, me assure you that I have done more stress and durability analysis than the usual Internet, Armchair Expert.
What is baffling about this fracture is that there are no visible, abrupt changes in geometry that say "stress riser". Nor, looking at the location in the overall structure, is there any reason to suspect high stresses at this particular point.
Perhaps there are internal gussets and reinforcements that are not visible but that would not be an economical place to put them from a fabrication standpoint.
While it is remotely possible that the beam was too short and a bit was welded onto the end to make it longer, the fracture is not quite perpendicular to the section. That makes the fractured weld hypothesis very unlikely.
What no responsible engineer wants to suggest is the possibility that steel beams with very-low, cold temperature impact toughness were built into critical, US infrastructure like bridges.
Low-carbon steel is a marvelous material. It is inexpensive, tough, amenable to common fabrication techniques and equally strong in all directions.
During WWII, steel became critical to the war effort. War runs on nitrates, oil, steel and blood.
Somebody had the bright idea adding a little bit of this and a little bit of that to common, low-carbon steel and increasing the strength by 50%. The thinking was that if it was 50% stronger than applications like ships would use 30% less.
Contributing to the soon-to-be-problem, production science suggested that welding was faster than riveting. Ships designed to transport US war materials to Europe were made of High-Strength-Low-Alloy (HSLA) steel and were primarily of welded construction.
U-boat predations of merchant shipping made more northerly routes advisable. The high seas of northern latitudes made it more difficult for Nazi U-boats to get a good periscope fix on targets to launch torpedoes.
Toilet paper comes in rolls. The strip of paper is scored with a row of tiny holes about every four inches. I trust that most of my readers have seen pictures of toilet paper if they have not personally used it.
The user unrolls the desired amount and initiates a tear at the closest row of perforations.
Almost always, the paper separates at the desired location. It happens with such reliability that those of us who use toilet paper rarely give it a thought.
Some materials are supremely insensitive to suggestions like the row of perforations. Bubblegum is a prime example.
Other materials are supremely sensitive to suggestions. Glass for instance. The slightest scoring of the surface will cause a pane of glass to fracture in the indicated location.
Back to WWII
It was well understood that higher strength steels were often more brittle than lower strength steels.
Alloys were prepared. Tests were run. Ships were built.
What nobody had anticipated was that the first generation of HSLA steel used to build the Liberty Ships were temperature sensitive. Alloys that performed well at 60F could fail miserably at 20F.
North Atlantic. Cold. Stormy. High, humped waves.
All welded construction.
Square-cornered access to holds, designed to facilitate rapid production.
Documented cases of captains looking for U-boat periscopes seeing six Liberty ship in one direction and by the time they swept back in that direction there were only five. The sixth ship had snapped in half and sunk just that quickly.
Why did ship-designers choose the first HSLA alloys?
The entire point of choosing them was that they didn't use rare, costly alloying element nor did they require exotic thermal processing at the mill. After all, mill capacity was a limiting factor.
Those economics have not changed. First generation HSLA steels are still less expensive. There is economic advantage to slipping a few (or more than a few) into the mix. After all, when the Interstate system was first built the ability to give every steel beam a unique bar-code label was a pipe-dream. It would not be that hard. And if they were sent to a hot state like Tennessee, what are the chances of them ever getting cold?
The crappy thing is...
The only good way to test for low-temperature notch sensitivity is to cut a sample, notch it, cool it and test it.
There might be twenty million steel beams built into bridges in the US.
Furthermore, those beams might be just fine until somebody comes along and cuts a sample out of them. Then the beam will be like the pane of glass with a line scored in the surface.
Like I wrote earlier, no responsible engineer wants this to be the case.
They really, really, really want to find some other "smoking gun". Otherwise it will be at total shit-storm.