I think this is useful even for systems (SW stacks) that are much smaller and "knowable": you start by observing, trying small things, observing more, trying different things, observe more and slowly build a mental model of what is likely happening and where.
His defining characteristic is where you can permanently work around a bug (not know it, but know _of_ it) vs find it, know it, fix it.
His defining characteristic is where you can permanently work around a bug (not know it, but know _of_ it) vs find it, know it, fix it.
Very interesting.