Engineering Hardships of Resuming Agents Midway After Crashes
benyounesarabah · reddit · 2026-07-06
The author analyzes the difficulty of recovering long-running agents after crashes (e.g., restarts, OOM, preemption): restarting duplicates side effects, while discarding leaves half-finished work. True resumption requires continuing from the breakpoint, but dependent states like conversation history and tool outputs vanish with the process.
The piece further discusses checkpoint granularity (per tool call, model call, or both), the latency impact of disk writes on hot paths, and the "silent divergence" issue during replays. It concludes that most frameworks still treat loops as ephemeral states fit only for demos, struggling to support production-grade long tasks.
More from coding & agent
- LangChain and Tavily to co-host NYC event on production agents using live web data, Nov 12 — LangChain · 2026-10-26
- SentinelForge: evidence-driven AI software repair with human-in-the-loop verification — Byte_Forge_2811 · 2026-10-26
- Dev hacks reMarkable e-ink tablet to let Opus draw and animate — Stefania_druga · 2026-10-26
- Godot games are trivially decompilable — compile C# to native DLL for real protection — mflux · 2026-10-26
- Techhalla shares full 2D-to-3D workflow with Opus 5.5 and Kling 4.0 prompts — techhalla · 2026-10-26
- Jason Furman: Building Econvitals pre-AI would need ~50 people, 3 years, $20M — HarrySurden · 2026-10-26