TL;DR (or Executive Summary, if you like): Between 620 and 1420 grams of mass (6.2 - 14.2 Newtons of force), depending on the pieces and their age.
So I decided to give this a shot and tried to measure the tensile strength of the technic chain links by building a homemade LEGO tensile tester machine and using chain link pieces from a new copy of 42107: Ducati Panigale V4 R for the simple chain and 42148: Snow Groomer for the tread elements - both received as part of the 2023 Q2 Activity Support initiative.
For the first attempt, I recreated the frame from my similar project with the train magnets, but a lot sturdier, as the chain links are obviously stronger than the magnets. The chain links were attached to the frame and the variable weight basket using gardening wire:
(pay no attention to the unfinished lower arches, they turned out to be totally unnecessary and I forgot to remove the partially built and still unconnected one before taking the photo)
Then I added batteries to the basket until the chain broke and measured the weight of the basket as it was at the moment...:
... at least, that was the plan. Sadly the basket was not heavy enough to break the chain even when filled to the brim and with additional batteries piled on the top. So a rework was necessary to accommodate a larger weight.
The next iteration included a 1.5-liter plastic bottle as the container for the variable weight, attached by the same wire loop to the chain link piece:
But with this approach, the measured breaking force would refer to the weakest of all the link pairs in the chain - while that could be an interesting measurement as well, I was mainly focused on one individual piece. Another rework resulted in the following contraption:
In this closeup, the chain attachment is visible along with the funnel used to fill the bottle gradually:
Using this setup the first real values were successfully acquired by filling the bottle until the chain broke, taking note of the weight displayed by the kitchen scale, replacing the link in the broken half, emptying the bottle and starting anew: (all values in grams)
If you noticed something, congratulations - I didn't, at least not at this point. So I turned to the sidequest: checking if the tracked link variant is measurably different.
Since those links are a lot wider, the attachment to the beam needed to be reworked again in order to access the chain for reattaching it after each breakage:
This setup has rendered the following measurements:
At this point I realized my mistake: after every breakage, I replaced the center piece of the three-link chain, that wasn't attached to the wire loops. But the end pieces were the same since the beginning, and thus were subjected to the wear and fatigue from countless measurements and failed tests. Their interfaces began to wear down, as it is (at least in hindsight) obvious from the first published data sequence, whose elements do indeed show a steadily declining trend:
So for the final rework, the wire loops became a lot simpler and messier, to facilitate the quick exchange of both participating links (only two this time!) against new ones.
Using two brand-new link pieces for all the following measurements, these data were gathered with the simple pieces:
And with the track pieces:
In conclusion, the tracked pieces are on average really a bit stronger than the simple ones (1362 grams vs. 1160, for a difference of ~15%), but both lose a lot of their strength (up to half of the new value!) when repeatedly forced to their breaking point. Even though there is no visible deterioration and the tests certainly haven't led to any destructive failures (apart from a few drops of spilled water).
The potential difference in strength resulting from turning the links "inside out" as recommended by AaronD in the comments became irrelevant by the test setup which uses a straight stretch of chain. It could be interesting once the chain is wrapped around a gear, but here I tried to get the cleanest measurement with the least amount of variables.
The kitchen scale used still is only accurate to about ~10 grams.
In the end, only 5-5 measurements were made with each link type using the final testing protocol - more measurements could refine the average and might introduce new extremes.