I'm sure Wikipedia used to have a much better write up, and more info on other clocks of a similar type. It's a shame they started on deleting and merging so much of the historic and esoterica - they lost a lot of accuracy and good data in the process.
Now the best write up and photo is at http://www.douglas-self.com/MUSEUM/POWER/oddclocks/oddclocks...
A printed volume has limitations that do not exist in a database or a filesystem. A wiki encyclopedia can easily preserve all information ever added to it, and still present it through as many interfaces as it likes.
If articles are tagged with enough metadata, it would not be extraordinarily difficult to produce all of the following automatically, with no further human intervention: abridged dictionary, unabridged dictionary, single-volume encyclopedia, multi-volume encyclopedia, single-disc encyclopedia, all math and science, all history, all art, all art minus the webcomics, all meaningful human knowledge, every last bit of trivia ever recorded, and everything written by anyone named Steven or Stephen.
It could be like moderating an entire library with the Slashdot system and then setting the default browse level to +4 or +5. But instead, the moderator-tyrants delete anything that they believe to be (-1, unimportant), and they can't be effectively metamoderated by the community.
Please feel free to add the info though, if you can find verifiable/reliable source to cite for it.
It's long enough ago I can't remember the exact route through that rabbit hole. Nonetheless the other point stands that much has been trimmed that seems like should easily meet the bar of adequate notability.
(At least, that's my impression. You can gain your own impression by looking at history, e.g. for article-deletion, here's the discussion/debate log from a day in the middle of last year: https://en.wikipedia.org/wiki/Wikipedia:Articles_for_deletio... )
Unfortunately, stuff that someone wrote on Wikipedia by themselves (without reference to some other source) does tend to get deleted, even if the person who wrote happened to be an expert… but that's the goal of Wikipedia: to summarized published material, rather than to contain original material. (See the “No original research” policy page at https://en.wikipedia.org/wiki/Wikipedia:No_original_research )
This policy does mean there's less useful information on Wikipedia than there could potentially be, but there do exist people who make up stuff, so having uncited material is problematic.
And, as you say, no original research pretty much guarantees that if something hasn't been documented somewhere credible, at least online, it tends to get excluded.
That said, the system works reasonably a fair bit of the time. I'm not even sure that the biggest issues are the result of Wikipedia policies. It's more that no system of this type works well when there's controversy and that even articles that were originally written well tend to accumulate random additions, factoids, and changes that degrade them over time.
People have tried alternatives to address perceived Wikipedia shortcomings. Google Knol was one such example that didn't work out. 
However, the mechanism that makes the clock tick-tock once per second is, again, separate from the winding mechanism, and I don't see why you'd have to (as another poster commented), "choose 1".
The Clock of the Long Now operates on similar principles (temperature difference), but separates timekeeping from the display mechanisms in part to maintain accuracy.
At the least this clock could take the title of longest running 'fake' perpetual motion machine.
It probably doesn't have a winder, but that doesn't mean it can't be wound.
> While the clock has not been wound [...] it has stopped on a number of occasions.
> when the ambient temperature has not fluctuated sufficiently to absorb the requisite amount of energy, the clock will not function
...so it's been wound.
“However, after environmental parameters readjust, the clock begins operating again.”
This suggests a lack of manual intervention.
Analog clocks don't have much to do, but I'm glad we still use them.
A quick search netted me this: https://www.ablicinc.com/en/doc/datasheet/real_time_clock/S3...
They specify a typical current of 0.25 uA at 3V supply, that's 0.75 uW power. So an energy of 18 uWh over 24h.
Same ballpark, but does not move gears (just talks on the "phone" ;-)
I am unsure if we have enough of an energy budget to get us some display (e-ink likely) (and the MCU also needs some extra energy for the updates) though.