Another factor influencing flow would be statistical fluctuation. For example say a process slot has a time of 2m 55s. In reality (especially where humans are involved) there would be a variance. This could be small or large so for example on average the quality control step takes 2m 55s but varies depending on car type, complexity and on how many defects are found. One remedy in this case is to improve product quality, another might be different qa slots per car, another might be reducing the amount of checks but could increase likelihood of customer defects. Same applies to a robot but the variance may be quite small, say no more than a second or two. Also applies to raw material supplies, the arrival rate from suppliers may vary a little or a lot.
Interesting, and I suspect a fair point, nothing flows truly like clockwork, until elon gets his machine-only factory i guess… 
Hehe yeah theres a big push towards robots and cobiotics but i think we are some way from self sufficient robots and true ai
There is Variance in everything, usually measured and explained by average (e.g. average process time) and range/std deviation (e.g. shortest process run time 1 min - longest process run time 5 min, range 4 min)
Its everywhere and one of the big problems - tea break length, supplier lead time variance, quality of a weld, machine power, a human’s speed in operating a process, everything that contributes to the first operation on a production line to the last
Right now, those doors and windows are really tough to install, so they don’t need any slowdown, lol