[BUG] Serena MCP tools (activate_project / list_memories) extremely slow or time out

### Preflight Checklist

- [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code

### What's Wrong?

Calling the Serena MCP tools mcp__serena__activate_project and mcp__serena__list_memories in a
Claude Code session, both calls take an extremely long time. activate_project took more than
2 minutes and was automatically moved to a background task before eventually completing.
list_memories ran in the background and failed after exactly 240 seconds with:
"Tool execution timed out after 240 seconds."

### What Should Happen?

Both tools should return within a few seconds, or at least fail fast with useful progress
feedback, instead of silently hanging for minutes — especially on a directory that is not a
typical source-code project (mostly JSON/TSV/Markdown data files, no language server needed).

### Error Messages/Logs

```shell
mcp__serena__activate_project on /home/c3phur/.exegol/workspaces
-> still running after 121s, moved to background task kfb833nqp
-> eventually completed with:
Created and activated a new project with name 'workspaces' at /home/c3phur/.exegol/workspaces.
Active language servers: .
File encoding: utf-8.
Onboarding has not been performed yet.

mcp__serena__list_memories (background task kh98x9m22)
-> failed: Error executing tool list_memories: Tool execution timed out after 240 seconds.
```

### Steps to Reproduce

1. Open a Claude Code session with the Serena MCP server configured, working directory set to
a git repo that is data-heavy rather than a typical source-code project (in my case: ~90
subfolders of JSON/TSV/Markdown files tracking bug-bounty programs, no language project
structure at all).
2. Call mcp__serena__activate_project with that directory as `project`.
3. Observe the call run well past 2 minutes and get automatically moved to a background task.
4. Immediately after, call mcp__serena__list_memories.
5. Observe it fail after 240 seconds with "Tool execution timed out after 240 seconds."

### Claude Model

Sonnet (default)

### Is this a regression?

I don't know

### Last Working Version

_No response_

### Claude Code Version

2.1.276 (Claude Code)

### Platform

Anthropic API

### Operating System

Ubuntu/Debian Linux

### Terminal/Shell

Other

### Additional Information

Working directory is a bug-bounty program tracking workspace (git repo), ~90 subfolders each
holding some_useful_files.json + markdown triage reports — not a software project. This
may explain why Serena's project indexing is unusually slow here. The calls were triggered
indirectly via the SuperClaude framework's /sc:save command, which defaults to Serena MCP for
session persistence.

Start by reproducing mcp__serena__activate_project and mcp__serena__list_memories in the data-heavy repository described, using the reported 121-second and 240-second behavior as reference. Trace the Serena MCP tool entry points and timeout handling; done means both tools complete within a few seconds or fail promptly with useful progress feedback.

Written by the indexing model from the issue text.