Remote-SSH Connection Pool Size Change in VSCodium 1.96 Causes Intermittent Exhaustion

Likely explanation

The intermittent ECONNREFUSED or timeout errors occur when the number of simultaneous SSH requests opened by VSCodium exceeds the current remote.SSH.connectionPool.maxSize limit. In VSCodium 1.96 the default limit was raised from 10 to 20, but many multi‑folder workflows (e.g., opening several projects plus active terminals) can still push concurrent SSH usage above 20, exhausting the pool. New requests are then queued or rejected until idle connections are returned, which is why the errors appear sporadically and disappear after a pause.

Confirmed facts

• VSCodium 1.96 introduced the setting remote.SSH.connectionPool.maxSize with a default value of 20.
• Changing this setting in settings.json does not take effect until the Remote‑SSH extension is reloaded (via the Command Palette → Developer: Reload Window or the Remote‑SSH indicator → Reload Extension).
• When the pool is exhausted, the Remote‑SSH log shows messages like “connection pool exhausted” and the client returns ECONNREFUSED or timeout errors.
• After idle periods, connections are returned to the pool and the errors subside.

Steps to resolve for this case

1. Estimate peak concurrent SSH usage: count the maximum number of folders you open at once plus any active remote terminals or tasks that spawn SSH connections.
2. Set an appropriate pool size: add or update "remote.SSH.connectionPool.maxSize": N in your user or workspace settings, where N is comfortably above your estimated peak (a common range is 15‑25).
3. Reload the Remote‑SSH extension: open the Command Palette (Ctrl+Shift+P or Cmd+Shift+P), run Developer: Reload Window, or click the Remote‑SSH indicator in the status bar and choose Reload Extension. This activates the new limit.
4. Verify the fix:
   • Open the Output panel, select Remote‑SSH log, and confirm that no “connection pool exhausted” messages appear while reproducing the load (e.g., opening the usual number of folders).
   • Optional: run a quick check of active SSH connections from VSCodium:

     # Find VSCodium PID
           pid=$(pgrep -f codium)
           # Count SSH sockets
           lsof -i -a -p $pid | grep ssh | wc -l
           

     Ensure the count stays below the configured maxSize during the test.

Missing diagnostic detail

To give a precise recommendation for maxSize, I need to know the observed peak number of simultaneous SSH operations (folders opened + active remote terminals/tasks) during the failure. If you can provide that count, I can suggest an exact value; otherwise, start with a value like 25 and adjust based on the verification steps above.