Well, 15m of tweeting on BART has led to dozens of high-value interactions, new followers, and insights for the upcoming blog post. Not bad as a case study!
My experience has shown that this is actually a much less stressful way to work. You're asking ppl to give estimates, with the caveat that there will be no penalty for not hitting it. Because all overruns will be absorbed in the common pool of safety margin, the buffer. This removes all the politics from estimating - the grandstanding, the making others look bad, the not-so-subtle pressures, the over-inflating to avoid looking bad. Estimates become what they're supposed to be: just estimates.
Also, notice that as hard as everyone says estimating is, they know EXACTLY when to start on that thing they're procrastinating on to finish it just before the deadline. Could it be this interesting phenomenon is a feature, not a bug?
But there is no correct time. That's a deterministic paradigm, whereas tasks are probabilistic events, with a distribution of possible points. Take a given task, and there are versions or scenarios where it takes 1x units of time, all the way to 100x. What critical chain does is help the worker play with the 3 sides of the PM triangle, time, budget, and scope, in a way that optimizes for the global throughput, not just their little local optimum.
There is a bit of a trick, but there are ways to make it easier. This is why in the tweetstorm I said the various parts of critical chain have to be used together, like forces balancing a building. In order to use median (most likely) time estimates and late starts, you also must use buffers. Instead of putting a little bit of safety margin at every little step, which explodes the lead time of the project w/out actually reducing risk (since delays are passed on and early finishes are wasted, as the table-building ex above explains), w/ buffers you pool all the safety margin together, where it can be used by whoever actually needs it. This works better because Murphy always strikes one part of the project disproportionately, not every step equally.
But what is the correct time? There is a whole subfield of project mgmt dedicated to exploring the theories around this question. And the line between "psychological games" and "psychology" is very thin, some would say non-existent.
Roadrunner mode probably sounds a lot like the "crisis mode" we're all used to in most orgs, but it's actually more like supercompensation principle in all types of learning and training. You focus w/ everything you have, and then recover. Most orgs try to balance capacity, which results in this steady state of draining yet not really productive work. Theory of Constraints (which is what this thread is based on) says you should instead balance flow.
OP here. Really cool to see the discussion here. So much deeper and more critical than anywhere else. I've made a note of many of the tools mentioned here and will check them out soon.