The current version of portage, 2.3.47, relies on multiprocessing.cpu_count() to determine the number of cores, which under LXD can give the number of cores on the host regardless of the actual # of CPUs allocated to the container. Portage should use a method that involves looking at /proc/cpuinfo instead.

The problem code is in lib/portage/util/cpuinfo.py.

Example difference:

# python3
Python 3.6.6 (default, Dec  1 2018, 08:39:55) 
[GCC 7.3.1 20181104] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import multiprocessing
>>> multiprocessing.cpu_count()
80

 # cat /proc/cpuinfo | grep "^processor" | wc -l
8