`node-test-pull-request` failing despite `node-test-commit` passing

We've got node-test-pull-request failing regardless of node-test-commit state. This is particularly problematic because the builds don't have a Resume and we can't land them with CQ nor can we resume the failing ones. Full CI re-run is the only option.

https://ci.nodejs.org/job/node-test-pull-request/77389/
https://ci.nodejs.org/job/node-test-pull-request/77385/
https://ci.nodejs.org/job/node-test-pull-request/77386/

10:20:24 quiet period for node-test-commit is 0 seconds.Finished Build : #92158 of Job : node-test-commit with status : SUCCESS at 04:20:24
10:20:24 Starting to gather test results!
10:20:24 FATAL: Channel "hudson.remoting.Channel@10050b49:JNLP4-connect connection from 67.158.54.159/67.158.54.159:61280": Remote call on JNLP4-connect connection from 67.158.54.159/67.158.54.159:61280 failed. The channel is closing down or has closed down
10:20:24 java.nio.channels.ClosedChannelException
10:20:24 	at org.jenkinsci.remoting.protocol.NetworkLayer.onRecvClosed(NetworkLayer.java:155)
10:20:24 	at org.jenkinsci.remoting.protocol.impl.NIONetworkLayer.ready(NIONetworkLayer.java:143)
10:20:24 	at org.jenkinsci.remoting.protocol.IOHub$OnReady.run(IOHub.java:808)
10:20:24 	at jenkins.util.ContextResettingExecutorService.lambda$wrap$0(ContextResettingExecutorService.java:26)
10:20:24 	at jenkins.security.ImpersonatingExecutorService.lambda$wrap$0(ImpersonatingExecutorService.java:66)
10:20:24 	at jenkins.util.ErrorLoggingExecutorService.lambda$wrap$0(ErrorLoggingExecutorService.java:51)
10:20:24 	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144)
10:20:24 	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642)
10:20:24 	at java.base/java.lang.Thread.run(Thread.java:1583)
10:20:24 Caused: hudson.remoting.ChannelClosedException: Channel "hudson.remoting.Channel@10050b49:JNLP4-connect connection from 67.158.54.159/67.158.54.159:61280": Remote call on JNLP4-connect connection from 67.158.54.159/67.158.54.159:61280 failed. The channel is closing down or has closed down
10:20:24 	at hudson.remoting.Channel.call(Channel.java:1101)
10:20:24 	at hudson.Launcher$RemoteLauncher.kill(Launcher.java:1162)
10:20:24 	at hudson.model.AbstractBuild$AbstractBuildExecution.run(AbstractBuild.java:538)
10:20:24 	at PluginClassLoader for jenkins-multijob-plugin//com.tikal.jenkins.plugins.multijob.MultiJobBuild$MultiJobRunnerImpl.run(MultiJobBuild.java:131)
10:20:24 	at hudson.model.Run.execute(Run.java:1849)
10:20:24 	at PluginClassLoader for jenkins-multijob-plugin//com.tikal.jenkins.plugins.multijob.MultiJobBuild.run(MultiJobBuild.java:61)
10:20:24 	at hudson.model.ResourceController.execute(ResourceController.java:97)
10:20:24 	at hudson.model.Executor.run(Executor.java:456)
10:20:24 Collecting metadata...
10:20:24 Metadata collection done.
10:20:24 [Checks API] No suitable checks publisher found.
Notifying upstream projects of job completion
Finished: FAILURE

Start with the linked node-test-pull-request builds 77389, 77385, and 77386, then compare their logs with the successful node-test-commit step. Investigate the JNLP4 channel closure while gathering test results; done means pull-request builds can complete result collection and support CQ or resume without requiring a full rerun.