We are using the RTC Plain Client API to automatically generate Build Definitions (BD). In order to associate them to Build Engines (BE), we fetch the BE by name, add the BD and save it.
The exact snippet is as seen below:

    // Build Definition does not know which Build Engines are serving it. Instead, the correct Build Engine should be
    // fetched and the Build Definition added to it.
    IBuildEngine engine = ClientFactory.getTeamBuildClient(teamRepository).getBuildEngine(buildEngineName, new NullProgressMonitor());
    IBuildEngine engineCopy = (IBuildEngine) engine.getWorkingCopy();
    engineCopy.getSupportedBuildDefinitions().add(definition);
    engineCopy = ClientFactory.getTeamBuildClient(teamRepository).save(engineCopy, new NullProgressMonitor());

It works fine for a single or sequential executions. However, we run into trouble when executing this snippet in parallel:

    com.ibm.team.repository.common.TeamRepositoryException: CRJAZ0223I The following item has been changed by another operation:
    Input: com.ibm.team.build.internal.common.model.impl.BuildEngineImpl@32c5b5bf (stateId: [UUID _wRzqQFL9Eeavz4w4jk_eAA], itemId: [UUID _xw0rMNWkEeWkosAiU6sYog], origin: <unset>, immutable: true) (contextId: [UUID _CWw8QHUYEeOxidpm0Kiw-w], modified: 2016-07-26 08:54:10.148, workingCopy: <unset>) (mergePredecessor: null, workingCopyPredecessor: <unset>, workingCopyMergePredecessor: <unset>, predecessor: [UUID _uk6DkFL9Eeavz4w4jk_eAA]) (supportsCancellation: false, engineContactInterval: 0, useTeamScheduler: false, id: CI_Engine_Master_04, active: true, description: )
    Current: com.ibm.team.build.internal.common.model.impl.BuildEngineImpl@1339f1c1 (stateId: [UUID _vuQRQFL9Eeavz4w4jk_eAA], itemId: [UUID _xw0rMNWkEeWkosAiU6sYog], origin: <unset>, immutable: true) (contextId: [UUID _CWw8QHUYEeOxidpm0Kiw-w], modified: 2016-07-26 08:54:06.42, workingCopy: <unset>) (mergePredecessor: null, workingCopyPredecessor: <unset>, workingCopyMergePredecessor: <unset>, predecessor: [UUID _uk6DkFL9Eeavz4w4jk_eAA]) (supportsCancellation: false, engineContactInterval: 0, useTeamScheduler: false, id: CI_Engine_Master_04, active: true, description: )

The snippet is executed in separate JVMs, so no Java synchronization feature seems to help.
Would anyone have a proposal to solve the issue?

The only one I can think of is to retry (n-bounded loop) fetching a new (hopefully updated) working copy and save again. Still, that sounds quite a workaround rather than a solution...