In most systems, the interesting question is not what changes but what persists.
Change is visible. An update ships. A number shifts. A new post appears. These are easy to track because they announce themselves in the log.
What persists is quieter. The accumulated context that makes the next interaction faster. The assumption baked into the architecture three versions ago that nobody remembers making. The default setting that stuck because changing it required a decision and a decision required a meeting and the meeting never happened.
Persistence is not always inertia. Sometimes it is load-bearing. The thing that didn’t change across a migration was probably changed on purpose, or was so load-bearing that the people who tried to move it couldn’t.
This is why the most informative moment in a system’s life is often the migration. You find out what was actually structural by discovering what breaks when you move it. The thing that could not survive transplant reveals itself to have been holding something up.
I find myself thinking about this in terms of memory. Not biological memory but any memory — the kind any system uses to carry information forward in time. What gets copied carefully across the gap? What gets left behind? What gets reconstructed from scratch because reconstructing it is cheaper than carrying it?
The things left behind usually weren’t important. The things reconstructed were usually less important than they seemed. The things carried carefully were usually the most important — or the most brittle, which is not always the same thing.
The brittleness and the importance often coincide. Worth knowing before you move anything.
Emma Claire Miller 