there is already a similar report at kernel bugzilla, but nobody took notice:
https://bugzilla.kernel.org/show_bug.cgi?id=190971

It's a kernel bug, first seen with linux-4.20 on my Carrizo. I don't have a
testcase for reproducing, so it's hard to bisect. Sometimes it happens multiple
times a day, with every suspend-resume cycle. Sometimes the machine works for
several days with at least two suspend-resume-cycles per day.