The problem is the following: On Internet Explorer or Firefox, when I load the deployment pipeline view, everything is connected correctly. You can see this in the screenshot "noscroll.png".
If I scroll the page, at first, everything is still connected correctly, that is, the arrows scroll along with the page.
However, after the update period expires, the arrows move so they go into the same absolute screen position in which they were before, which is not the correct position, since the cells in the pipeline have moved, so they are in a different absolute position. You can see this in the screenshot "scroll.png".
In informal language, if I point a finger to an arrow in the screen, then scroll, the arrows stay fixed where my finger is, but the pipeline cells move according to my scroll.
The problem has been present since I installed the plugin recently. I'm using the latest available version of both Jenkins and this plugin.
The problem can NOT be reproduced in full screen view. This works correctly.
This problem can not be reproduced in Chrome.
The problem seems superficially similar to the following problem: https://groups.google.com/forum/#!topic/jsplumb/D2ghz7tV4m8
This is the full list of plugins I have installed (not including dependencies). All updated to the latest version:
Copy Artifact Plugin 1.37
CVS Plug-in 2.12
Delivery Pipeline Plugin 0.9.8
Email Extension Plugin 2.41.3
Job Generator 1.22
MSBuild Plugin 1.25
MSTest plugin 0.19
Nested View Plugin 1.14
Node and Label parameter plugin 1.7.1
Node Iterator API Plugin 1.5
Role-based Authorization Strategy 2.2.0
Self-Organizing Swarm Plug-in Modules 2.0
SonarQube Plugin 2.3
SSH Slaves plugin 1.10
Throttle Concurrent Builds Plug-in 1.8.4
Translation Assistance plugin 1.12
VSTest Runner plugin 1.0.4
NOTE: The delivery pipeline view I am creating is NOT inside a nested view.
If I can gather enough time, I will set up a clean install to see if I can reproduce the problem, and install the plugins one by one to see if one breaks the page, since I expect that this issue would already be reported if it happened on a common setup.