Reference next dev cache corruption on restart

During the CreatorTrack 2026-07-03 live-iteration session, ~25 rapid systemctl restart cycles on the next dev process corrupted Turbopack's incremental cache (.next-dev). Symptom: pages served fine (/login 200) but all /api/auth/* routes (providers, session, csrf) 404'd, breaking the whole site via NextAuth failures. The route file and middleware were unchanged and unrelated commits were ruled out before landing on cache corruption as root cause.

Why: Restarting a dev server repeatedly under load can leave its incremental build cache in a state where route handlers stop registering, without any code change being at fault. This wastes debugging time chasing phantom code regressions.

How to apply: If next dev starts 404ing routes it clearly serves in code, don't keep restarting. Do a clean rebuild: rm -rf .next-dev and cold-recompile, then verify auth/API routes return 200 before assuming a code bug. This is now standard practice after any batch of dev restarts on CreatorTrack, per the merge loop (worktree agents -> merge -> clean-rebuild dev, never bare-restart) documented in the 2026-07-03 handoff.

[auto-memory session c13c7aac-6db6-4f78-86c1-601ee921164b, confidence 0.85, mode direct]