Cloud Image can't be moved to a pool not containing the project that contains this image.
Dorm: Canceled build without agent cannot be displayed: "You do not have enough permissions to view requested agent pool".
.NET builds with qualifier agent requirements starting with "Exists=>" cannot find compatible agents if there are incompatible agents in the same pool.
Lots of duplicate entry errors on attempt to store a metric id to the dictionary table.
Perforce failing on newlines in changelist description.
Agent alternate IP address ignored by TeamCity.
Working better with non-default branches using MCP.
Cleanup of build type groups can fail with an unknown error.
Rake build steps are broken (plugin is damaged).
Authorize virtual agents action should validate agent pool for cloud agents.
[S3] Uploads fail with 403 status code.
Build-scoped tokens UX: add descriptions and docs link.
Properties for key preserving on rotation (time.min and time.days) are not working.
TeamCity cleanup does not remove cancelled multi-node tasks.
Creation flow: handle Unauthorized exception in the UI.
Excessive '/user' calls during repository listing in BitBucket Cloud connection.
Provide more information about the origin of the TeamCity server workspaces.
IndexOutOfBoundsException in org.jetbrains.teamcity.fus.client.events.TeamCityListener.
Mark test ignored if it was interrupted and parent block gets closed.
Messages "Agent will upgrade" and "Agent can't be upgraded automatically" can be shown at the same time.
Incorrect "mcpClientToolName" sent to FUS during connection to MCP with Gemini, Copilot and some other tools.
Project import loads the whole custom_data_body table.
Saving settings on Windows can be slow.
Repeating changes collection operations holds a reference to a previous operation via '$0'.
Add a primary key to the ignored_tests table to help improve performance of DB replication.
BuildPTRIndexer$BuildTask.beforeExecute consumes CPU in case of a large queue.
PageExtensionsInterceptor is called for non-WEB POST requests.
Calculation of unprocessed node events takes a lot of time.
Processing of the events blocks on VCS commits cache initialization.
Slow loading of VCS leads to UI slowness.
Server cleanup duration ~3x regression on SQL Server after 2025.11.3 upgrade.
Too big thread name with lots of "commit" strings.
Duplicated IDs in the where clause of the SQL query can cause the PostgreSQL planner to switch to a full table scan.
Avoid loading the entire VCS root DAG of commits in memory.