# Intermittent User Search Failures (Jan 2026)

| **Severity**   | P1             |
| -------------- | -------------- |
| **Status**     | REsolved       |
| **Start Date** | Jan 22, 2026   |
| **End Date**   | Feb 4, 2026    |
| **Duration**   | 13d            |
| **Impact**     | 39 users       |
| **Tickets**    | ECOHELP-108052 |

---

## [#](#overview)Overview

'Capable for Confluence' and 'Approval for Confluence' are impacted by intermittent Confluence User Search API failures, causing unreliable user searches. Atlassian acknowledges this as P1 but has not posted publicly or provided a timeline for resolution.

---

## [#](#incident-description)Incident Description

Calls to the Confluence API `/wiki/rest/api/search/user` endpoints fail with `` 500 Internal Server Error: "UndeclaredThrowableException"` or async search exceptions ``.

This stops users from searching for and adding others to an approval request. Sometimes it works; other times, you get the error shown below.

This affects a subset of users and may be related to how many groups or sites a specific user has.

---

## [#](#how-do-i-know-if-i-m-affected)How do I know if I’m affected?

![No results for user search](https://help.gocapable.com/images/9529a5ba-b781-440b-88b0-2f49efad0f84.webp)

User Search returns no results

![User search incident message](https://help.gocapable.com/images/622ea740-ca4e-4728-a9bd-778c612cd3ca.webp)

Error notification is displayed

---

## [#](#workaround)Workaround

We've investigated alternative APIs like `/wiki/rest/api/search/user` and GraphQL betas, but none fully match the dynamic CQL user querying needed in Forge apps for Capable Approval.

This intermittent Confluence Search API issue (tied to ongoing CQL limitations and spikes) often resolves with a 5-10 minute retry.

As a workaround, try inserting via `/approval` macro or adding approvers directly in the editor - some users report success bypassing the search dialog.

---

## [#](#incident-timeline)Incident Timeline

| **Date**     | **Time**                  | **Event**                                                                                                         | **Party** |
| ------------ | ------------------------- | ----------------------------------------------------------------------------------------------------------------- | --------- |
| Jan 22, 2026 | 12:34                     | Initial report of P1 incident                                                                                     | Capable   |
| 13:19        | P1 Confirmed by Atlassian | atlassian                                                                                                         |           |
| Jan 23, 2026 | 10.50                     | Requested progress report, ETA and public acknowledgement on [status.atlassian.com](http://status.atlassian.com). | Capable   |
| 12.48        | Escalated issue.          | Capable                                                                                                           |           |
| Jan 26, 2026 | 09.00                     | Added in-app messaging when issue occurs.                                                                         | Capable   |
| Jan 28, 2026 | 17.00                     | Investigation in progress.                                                                                        | Atlassian |
| Feb 2, 2026  | 09.30                     | Reproduction app provided                                                                                         | Capable   |
| Feb 4, 2026  | 02.00                     | Patch is ready for release                                                                                        | Atlassian |
| Feb 6, 2026  | 06.00                     | Issue is confirmed as closed by multiple customers                                                                | Capable   |

---

## [#](#atlassian-analysis)Atlassian Analysis

The root cause was related to the number of scopes defined in Forge manifest.

However, this didn’t directly cause the issue. The issue happened in between two internal services which talk to each other to serve the API requests.

These services user their own Authorization headers (not exactly the one you’d make the API call with) which can increase with the context size. And, the context size can increase with the number of scopes a Forge app has.

A proxy between the two services had a smaller header size limit - which caused the issue.