Codex Logs Benchmark SSD Mortality at 640 TB

Description

A cropped GitHub issue on a white page shows the header "1996fanrui opened last week" and the bold title "Codex SQLite feedback logs can write ~640 TB/year and rapidly consume SSD endurance." Beneath an "Issue" heading, the report says Codex continuously writes large amounts of data to its local SQLite feedback-log database and lists `~/.codex/logs_2.sqlite`, `~/.codex/logs_2.sqlite-wal`, and `~/.codex/logs_2.sqlite-shm`. The remaining text says that after about 21 days of uptime the machine's main SSD had received about 37 TB of writes, with process/file-level checks identifying the Codex SQLite logs as the main continuous writer. It extrapolates that rate to roughly 640 TB per year—about 640 full-drive writes on a 1 TB SSD—and warns that a consumer drive rated near 600 TBW could exhaust its warranted endurance in under a year, making excessive diagnostic logging and SQLite WAL churn the darkly comic failure mode.