Glitch boot failure diagnosis when console logs are truncated

Diagnosing a failed deployment on Glitch can be complicated when the container crashes repeatedly during the boot sequence. The integrated console logs capture stdout and stderr from the application process, but they are ephemeral and may be truncated or lost before the app starts. This is particularly relevant when the automatic sleep/wake mechanism triggers a cold start and the external caller times out before the container stabilizes.

Because free-tier sleep cycles cannot be disabled and certain system-level configurations reset on restart, the boundary between platform orchestration and user-visible logging is unclear. Are there documented limits to how much boot output is retained when a project fails to deploy, can failures occurring before the application process begins be diagnosed from the project console at all, and how does the sleep/wake cycle affect log retention during repeated boot failures?