Cursor hang on reload - corrupted state.vscdb

Support Help
,
1

Hi all,

Not sure if this is a bug, but after stopping and restarting the Cursor IDE several times today (Linux, v. 2.6.21 / x64) Cursor suddenly presented me with the first-run screen and required me to re-authenticate via the browser.

After doing so, it showed me the initial Preferences screen, the Welcome to Cursor modal, and then hangs on a Loading… screen (after very briefly showing the usual Cursor UI for a few milliseconds). It appears that the state.vscdb file has been corrupted:

-rw-r–r-- 1 cursor 188416 Mar 26 11:10 state.vscdb
-rw-r–r-- 1 cursor 188416 Mar 26 11:10 state.vscdb.backup
-rw-r–r-- 1 cursor 16994308096 Mar 26 11:04 state.vscdb.corrupted.1774548302373
-rw-rw-r-- 1 cursor 20 Mar 26 11:14 state.vscdb.options.json
-rw-r–r-- 1 cursor 32768 Mar 26 11:15 state.vscdb-shm
-rw-r–r-- 1 cursor 131872 Mar 26 11:15 state.vscdb-wal

I’ve been actively developing a cursor plugin, so had been enabling/disabling/reconfiguring various things in the top-level mcp.json and local .cursor/plugins/local directory, but these do not seem like actions that should’ve caused this.

After trying to stop / restart several times more (invoking ‘cursor’ via the bash command line), I noticed the following entry in my /var/log/syslog –

2026-03-26T11:22:42.223659-07:00 surface-mint systemd[4187]: Started vte-spawn-511cad45-62b3-4488-a7a0-6c67886a7074.scope - VTE child process 42906 launched by gnome-terminal-server process 41486.
2026-03-26T11:22:45.333621-07:00 surface-mint systemd[4187]: Started app-org.chromium.Chromium-42932.scope.
2026-03-26T11:22:49.910278-07:00 surface-mint kernel: audit: type=1400 audit(1774549369.909:164): apparmor=“DENIED” operation=“bind” class=“net” profile=“cursor_sandbox” pid=43069 comm=“cursorsandbox” family=“netlink” sock_type=“raw” protocol=0 requested=“bind” denied=“bind”

I was eventually able to get to the Project screen, and after selecting my remote SSH project workspace, my project loaded (albeit in Agent mode, not Editor mode), with a warning banner ‘Restricted mode is intended for safe code browsing. Trust this folder to enable all features’.

In the Window output log, I also see a lot of repeated errors (can share full log if needed) like

2026-03-26 11:17:24.344 [error] [Window] No Connect transport provider registered.: Error: No Connect transport provider registered.
at Nfa.transport (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:34180:614573)
at async ZS.createSingleServer (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:34180:616165) No Connect transport provider registered.
2026-03-26 11:17:25.137 [error] [Window] No Connect transport provider registered.: Error: No Connect transport provider registered.
at Nfa.transport (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:34180:614573)
at async ZS.createSingleServer (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:34180:616165)
at async ZS.get (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:34180:615371)
at async Zrd.keepCppModelStateUpdated (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49822:16507)
at async Zrd.loadCppConfigIncludingHandlingProAccess (vscode-file://vscode-app/usr/share/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49832:1259) No Connect transport provider registered.

Has anyone else encountered this? I am thoroughly confused.

edit: To try to get things working, I moved the old corrupt .vscdb files out of globalStorage, but had to explicitly trust my workspace directory via this dialog, which I’ve never seen before. Is this a new thing?

Screenshot from 2026-03-26 14-19-06
Screenshot from 2026-03-26 14-19-061007×599 53 KB

After last update, stuck at "Waiting for extension host"
3

Hey, you’ve run into two well-known issues at the same time. I’ll go through them in order.

  1. Corrupted state.vscdb around 16 GB

This is a known bug. state.vscdb can grow out of control. Your corrupted file is 16 GB, which matches what we’ve seen from other users. The team is aware.

Related threads:

What you did, moving the corrupted files out, is the right workaround. Cursor created a new state.vscdb 188 KB, which is a normal size. The side effect is that you lose some settings, including trusted folders, so you’ll see the Workspace Trust dialog. That’s expected behavior, not a new feature.

  1. AppArmor denial cursor_sandbox

This is also a known issue. The AppArmor profile shipped with Cursor is incomplete. It’s missing network rules and some other permissions. A detailed breakdown and a workaround are here: AppArmor cursor-sandbox profile incomplete on Linux — sandbox does not work correctly (missing network, signal, userns)

Quick but less safe option:

sudo sysctl -w kernel.apparmor_restrict_unprivileged_userns=0

Better option is to patch the AppArmor profile manually using the instructions in that thread.

  1. No Connect transport provider registered

This error was most likely caused by the corrupted state combined with an SSH remote workspace. After recreating state.vscdb, it should go away. If it still happens, let me know.

One question: is the project stored on NFS or another network filesystem? That can be a cause of corruption.