Hi all,
Registering a persisted Cypher query fails every time for us with a 502, including the example from the documentation. Validation clearly runs, but the store step fails. Has anyone else seen this since around 5 November, or is it something on our account?
What we send (from the logged-in browser session of the site admin, as the persisted queries guide and others describe):
POST https://one-atlas-xqiu.atlassian.net/gateway/api/twg/cypher-queries
Content-Type: application/json
{
"appId": "ari:cloud:ecosystem::app/a6f981fc-0c15-45e9-9f97-131478848a6b",
"name": "doc-example",
"cypher": "MATCH (user:AtlassianUser {ari: $userId})-[:atlassian_user_assigned_jira_work_item]->(item:JiraWorkItem) RETURN item",
"paramsSchema": [{ "name": "userId", "type": "ARI", "required": true }]
}
What we get:
HTTP 502
{"errors":[{"code":"BAD_GATEWAY","message":"Unable to store Cypher query"}]}
atl-traceid: 4a22d974268240e6a7170dec2d5c8542
x-failure-category: FAILURE_ORIGIN
date: Tue, 10 Nov 2026 21:19:33 GMT
What we ruled out
- The query. The documentation’s own example fails. So does a query we confirmed returns the right rows in TWG Labs (
jira_work_item_blocks_jira_work_item). - Validation and auth. A deliberately invalid query gets a proper 400
UNSUPPORTED_OPERATION(“Bidirectional relationships are not supported”, tracef81943f02d944d92a2b0dfda5af91d7c). An app ID we don’t own fails earlier with 502 “Unable to verify Forge app ownership”. Ours passes both and fails only at “Unable to store”. - The client. TWG Labs’ own Save query button (TWG Apps) sends the same request and gets the same 502. The dialog closes without an error and nothing appears under Saved queries.
- The site. It fails the same on a brand-new developer site (
ripple-geist.atlassian.net, created today). - The app. It fails for two separately registered apps (
a6f981fc-0c15-45e9-9f97-131478848a6band6aa27d28-8562-408c-a80f-c5d63b5d8bb4), and for the first one in both the development and production environments. The first app hasread:graph:issue:jiraandread:graph:user:atlassianand is installed in Jira and Confluence; the second is a plain Jira app without graph scopes. - Timing. It failed every time we tried over about two hours on 10 Nov 2026, including retries a few seconds apart.
For comparison: the same request pattern (browser session, same body, credentials: 'include') returned 201 for another Codegeist 2026 team on 5 Nov 2026 (handoff-brief on GitHub). That makes us think either something changed after 5 Nov, or it’s specific to our account or developer space.
Everything else works: curated queries (with X-Query-Context), requestTeamworkGraph from a Forge Rovo agent, and running the same Cypher in TWG Labs all return data.
Any pointers, or a confirmation that this is a known issue, would help a lot. Thanks!