nexus-staging-maven-plugin inline deploys perform slower than maven-deploy-plugin even when not using staging features

Builds using skipLocalStaging option and SNAPSHOT artifacts can take up to 2x longer using nexus-staging-maven-plugin as compared to maven-deploy-plugin.

See README.md in attached reproduce case.

Excerpt:

Using Maven 3.2.1:

Setup: Prime local repo

./mvn clean install -f build1/pom.xml

Setup: Local Nexus instance

Start Nexus at http://localhost:8081/nexus

Execute build using maven-deploy-plugin, no deploy at end. Observe wall clock build time reported by Maven.

time verbose=y bash -x ./mvn clean deploy -f build1/pom.xml -Dbuildsupport-staging.skip -T 4

Execute build using nexus-staging plugin, no deploy at end. Observe wall clock build time reported by Maven.

time verbose=y bash -x ./mvn clean deploy -f build1/pom.xml -DskipLocalStaging=true -T 4

Problems?

• nexus-staging plugin is usually 2x slower, but sometimes just as fast.
• total per module build time != total wall clock time as there are parallel threads executing (expected but maven-deploy-plugin does not seem to have this problem)
• jdk8 has a long delay at end of build doing some sort of cleanup, jdk7 always ends immediately ( just to note in case someone is looking for general speedup)
• for every `[INFO] org.sonatype.nexus.maven.staging.deploy.DeployMojo - Waiting for other projects build to finish... `
  log statement, NSMP adds an artificial delay of 2 seconds

Expected:

• why wait for other modules at all if we are deploying SNAPSHOT versions and not using deploy at end or staging features?
• "the correct way to do this in m3 is using LifecycleParticipant"