We are building a regression test using the matrixbuild plugin with one user-defined axis (17 values).
The build is triggered by an svn-serverhook upon new commits to the repository. Recently a user did batchcommit 23-commits using gitsvn, which triggered 23*17=319 Builds via jenkins in rapid succession.
220 of those 391 Builds died with an Nullpointer exception in net.praqma.jenkins.plugin.reloaded.Util.getDownstreamRebuildActionFromMatrixBuild (See the attached build.log for a detailed backtrace).

Building the project separately, all by it self, triggering it just once works fine, the bug seems to be connected to the number of concurrent builds.

To check this hypotheses we fired a series of concurrent buildjobs for the same matrix-build project (see steps to reproduce below). In contrast to the
builds triggered by the commits above the code remained constant.
This time, some builds encountered the NPE halfway through the build, so the backtrace differed to some degree, but the last few invokes in the stacktraces kept the same as before.
(See the attached build_reproduced.log for a full backtrace)

The issue is reproduceable:
Steps to reproduce:

• For said matrix-build-project (most likely any matrix-build project (even though this was not tested)) get the token from the configuration page.
• Query the https://<jenkins-installation>/job/<project>/build?token= multiple times in short order. 10+ subsequent requests seem to trigger the issue reliably

As the number of concurrent builds has influence on the issue and we did not really have those before, I can't say anything about older versions of jenkins or the respective plugins.

I'm not 100% sure about the priority, but as the description for "critical" explicitly mentions crashes. If this categorisation does not match the common practice for this bugtracker, please reclassify.

I would be happy to provide further information if necessary, just ask.

Regards,
Simon