fix: consolidate memory leak fixes from #14650 and #8953

[tda1017]

Issue for this PR

Closes #9385
Relates to #7046, #9151

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

This PR consolidates memory leak fixes from two existing PRs that have been open for a while:

• fix: resolve memory leak issues across multiple subsystems #14650 by @kryptobaseddev
• feat(bash): stream large output to tmpfile with filtering and GC #8953 by @cgwalters

All credit goes to the original authors. I'm submitting this to help move these fixes forward.

Key fixes included:

1. Subagent session deallocation (tool/task.ts) — Calls Session.remove() after task completion. Previously subagent sessions were never cleaned up, causing GB-scale memory growth.

2. Delta string cap (event-reducer.ts) — Caps accumulated part field strings at 1MB with FIFO eviction.

3. Bash output streaming (tool/bash.ts) — Streams command output to tmpfile when >50KB, preventing O(n²) memory growth for large outputs.

4. SSE buffer guard (serverSentEvents.gen.ts) — 10MB limit prevents unbounded growth during stalled streams.

5. MCP OAuth transport TTL (mcp/index.ts) — 5-minute auto-eviction for stale OAuth entries.

6. TUI event listener cleanup (app.tsx, session/index.tsx, prompt/index.tsx) — Properly unsubscribes 8 event listeners on component unmount.

7. LSP diagnostics cap (lsp/client.ts) — 2,000-file cap with FIFO eviction.

8. Additional — AsyncQueue termination, bus cleanup, PTY chunking, shutdown disposal.

How did you verify your code works?

Built locally and ran extended sessions. The original PRs (#14650, #8953) contain detailed testing methodology — @kryptobaseddev ran 2.5 hour stress tests showing RSS stabilized at ~2.5GB vs previous 25GB+ growth.

Screenshots / recordings

N/A - backend memory fixes, no UI changes.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

---

Co-authored-by: Keaton Hoskins kryptobaseddev@users.noreply.github.com
Co-authored-by: Colin Walters walters@verbum.org

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Related PRs Found

The current PR (#15435) is a consolidation of two existing PRs that address specific memory leak issues:

Primary Sources (being consolidated):

• fix: resolve memory leak issues across multiple subsystems #14650: "fix: resolve memory leak issues across multiple subsystems" - by @kryptobaseddev (10 memory leak sources)
• feat(bash): stream large output to tmpfile with filtering and GC #8953: "feat(bash): stream large output to tmpfile with filtering and GC" - by @cgwalters (bash output streaming)

Related Memory Leak Fixes (addressing similar issues):

• fix(web): dispose idle MCP server instances to prevent process accumulation #15326: "fix(web): dispose idle MCP server instances to prevent process accumulation"
• fix: clear tool output and attachments when pruning to prevent memory leak #7049: "fix: clear tool output and attachments when pruning to prevent memory leak"
• fix: add LRU eviction to LSP client file and diagnostics tracking #7050: "fix: add LRU eviction to LSP client file and diagnostics tracking"
• fix: clean up pending permissions when session is deleted #7154: "fix: clean up pending permissions when session is deleted"
• fix(core): add dispose functions to prevent subscription memory leaks #7032: "fix(core): add dispose functions to prevent subscription memory leaks"
• fix: resolve multiple memory leaks causing unbounded RAM growth ( Now passing: && Fixed failing tests for windows ) #13514: "fix: resolve multiple memory leaks causing unbounded RAM growth"

These related PRs address similar memory management concerns across different subsystems (MCP servers, LSP diagnostics, permissions cleanup, subscriptions). While #15435 is specifically consolidating #14650 and #8953, the other PRs represent parallel memory leak fix efforts that may warrant review for consistency or potential interaction effects.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍