I encounter plasma becoming unresponsive as plasmashell CPU usage grows to 100%. All input will then become very slow and the screen/plasma may start to flicker. I did not wait for it to crash, but I had to force reboot the system in order to commence work. The problem can be worked around by issuing kquitapp5 plasmashell; kstart plasmashell whenever the CPU usage of plasnmashell grows to unusual values. I am running KDE Neon user edition with all current updates and encounter the bug since a two-or-three plasma versions (since about half a year). I cannot say definitely what causes the bug despite I am running the same variant of KDE on two very similar notebooks, Lenovo Thinkpad T520 and T530 respectively. Only on latter the bug occurs. One major difference is that I use it for work and regularly connect it to external monitors. Indeed, it seems that connecting or deconnectiong a monitor (which will cause plasma to run on both screens / only a single screen, with my settings) triggers the issue. (Another difference is that I use the T520 with discrete nvidia graphics while the T530, having the problem, features only an on-chip Intel graphics card.) When I work only on the notebooks native screen, the issue does not appear. To automate the work-around, I described above, I have a custom script register the kquitapp5 plasmashell; kstart plasmashell commands to the event of screen-unlocking, using dbus. With this solution, I do in practice not encounter the problem of high CPU usage at all. However, I see the task bar building up again, each time I unlock, which is not what I would want. There is a few related bugs (e.g. https://bugs.kde.org/show_bug.cgi?id=312919 and https://bugs.kde.org/show_bug.cgi?id=356479), but I did not see the problem being related to the attachment of external screens.
High CPU is a vague symptoms with many causes throughout the stack. Some known and fixed, some known and not fixed, some presumably not known. See the "see also" in the linked report for active specific bugs. *** This bug has been marked as a duplicate of bug 378010 ***