To answer archaeologists' question, we've moved away from clockworks due primarily to market forces. Tiny gears, springs and pendulums (or wound springs for watches) are difficult to produce and just as difficult to repair. Circuit boards are incredibly cheap to make these days and after a few decades of work, can now be much more accurate than the best mechanical clock. They aren't always better since less accurate watches will always cost less to make, but who really cares if their watch gains or loses a few seconds a year? Further, with digital watches, features are vastly easier and cheaper to include -- a mechanical watch with a timer and a second time-zone setting would cost hundreds whereas it can be implemented in a digital watch for a couple bucks.
You can still buy mechanical watches and there are still some people who make their living building and repairing the mechanical insides, but they cater primarily to the rich and those with heirlooms since the vast majority of us prefer the cheapest with the most functions, digital dominates.
As to the OP, the program is genius and I wish the code was readable so I could play with it myself! The first thing that struck me was how few parameters were fixed. Unlike many other simulations, there are a number of different pieces that can all be copied and adjusted. I'm particularly curious how he represented the variable number of pieces in matrix form (more a coding curiosity than a conceptual issue).
What I'd really like to see would be a changing environment. I'd love to run this program on a cluster and have different populations evolve to different time standards (i.e. 70 seconds per minute rather than 60 etc...). In fact, if you produced multiple time standards in a single 'environment' you could even simulate overlapping niches and have multiple populations competing for some of the same resources. As an example, if you had 40 s/min and 90 min/hr as one niche, and then 80 s/min and 45 min/hr as a second niche, you could potentially get two different populations that were coexisting and yet still competing for the 1-hour 'resources.'
Getting the balance right so that one population didn't out-compete the other would be tricky, but unfortunately I think the biggest problem would be computation time as these simulations apparently took a few weeks.
Ultimately, if the environment (ratio of sec/min/hr) were shifted at slightly trending speeds, I would be curious to see if you could reproduce a significant increase in complexity. The actual design would become more important and preserving a varity of designs (i.e. different mutations non-coding DNA that can be activated later) would allow some populations to be more successful in changing environments. Of course more gears would also help with adaptation though I'm not sure if there's a biological analog to high complexity in the clocks that also increases adaptability.
