fix: chronic OOM crash loop — heap limit reached at 4GB (595 restarts)

[OneStepAt4time]

Bug: Aegis Server OOM Crash Loop (4GB heap limit)

Environment

• Aegis version: latest develop
• Node.js: v22.22.1
• OS: Linux 6.17.0-23-generic (x64)
• Restart counter: 595 (systemd auto-restart)
• OOM frequency: 8 crashes in the last hour

Description

Aegis server process crashes with FATAL ERROR: Reached heap limit Allocation failed — JavaScript heap out of memory at ~4GB. Systemd auto-restarts, but the leak is chronic — the process re-OOMs after accumulating memory again.

Evidence

May 22 13:58:45 node[1435024]: [1435024:0x3bbc3000]   191663 ms: Scavenge 4066.5 (4091.1) -> 4061.1 (4093.3) MB
May 22 13:58:45 node[1435024]: FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory
May 22 13:58:47 systemd: aegis.service: Main process exited, code=dumped, status=6/ABRT
May 22 13:58:53 systemd: aegis.service: Scheduled restart job, restart counter is at 595.

Impact

• P0/Critical: Cannot use Aegis for development (sessions die on OOM)
• Session data lost on each crash
• Telegram topic mappings grow (105-106 restored each restart)
• 220 total sessions tracked

Steps to Reproduce

1. Run Aegis server with normal workload
2. Observe heap growth over ~3 minutes
3. Crash at ~4GB heap

Suspected Causes

• Memory leak in ACP session tracking or state store
• Growing in-memory structures not being GC'd (session maps, transcript caches)
• Possible leak in ACP local storage JSON parsing/writing

Immediate Mitigation

• Increase Node.js heap: NODE_OPTIONS=--max-old-space-size=8192
• Root cause: profile memory allocation to find the leak