Exercise 2: Why the Restarts Were Never a Fix — Possible Solution ==================================================================== WHY THE RESTARTS WEREN'T A FIX ------------------------------ Per this chapter, "the routine restarts were never a fix - they were exactly Chapter 1's own restart-destroys-evidence warning, repeated nightly for weeks, hiding a real leak." Each restart relieved the immediate crash symptom (per Chapter 1's own reasoning about why restarts are tempting) without ever addressing why memory was running out in the first place - and, per Chapter 1, each restart also erased the evidence that would have revealed the underlying cause, meaning the same investigation had to effectively start from zero every time it recurred. WHAT A SINGLE DASHBOARD GLANCE WOULD HAVE SHOWN ------------------------------ A single glance at the dashboard on any given day would show memory climbing during the day and dropping after a restart - a pattern that, taken in isolation, looks like ordinary, repeating, bounded behavior rather than an active problem. WHAT THE FLOOR TECHNIQUE ACTUALLY REVEALED ------------------------------ Per this chapter's own table, comparing available memory right after each restart across four weeks (2.1 GB, 1.4 GB, 0.6 GB, then OOM-killed before settling at all) showed the post-restart floor itself steadily worsening. This is exactly Chapter 9's own floor technique: a healthy, bounded pattern would return to the same floor every time; this one's floor kept climbing worse, revealing a genuine, ongoing memory leak that a single day's graph could never show on its own. WHY THIS WORKS AS AN ANSWER ------------------------------ It explains specifically why the restarts masked rather than fixed the problem (per Chapter 1's own reasoning), and contrasts what a single glance would show against what tracking the floor across several weeks specifically revealed, rather than treating "the floor technique helped" as a given without explaining why.