PostgreSQL read acl optimizations crash under high write concurrency

If asynchronous event listeners create many new documents in the same folder concurrently, the call to CoreSession#save() fails randomly throwing the exception mentioned in the first comment of this jira issue. Note: care has been taken to ensure that the names (local path segments) of the newly created documents do not collide.

Adding:

<aclOptimizations enabled="false" />

inside the tag:

<repository name="default"
factory="org.nuxeo.ecm.core.storage.sql.ra.PoolingRepositoryFactory">
<repository name="default">
[...]
</repository>
</repository>

in the default-repository-config.xml file in a custom configuration template makes it possible to workaround the problem.