Thanks for sending those files. It’s actually a bug in GPlates - I’ve just fixed that now. You can download gplates_2.3.1-beta.7_win64.zip here.
Your workflow exposed the bug because you have such a large time range going back to 3800 Ma. And so, in the Kinematics Tool, there were greater than 512 reconstruction times in the graph (ie, (3800 - 0) / 5Myr). The number 512 is a general internal limit (not specific to the Kinematics Tool) that we use to ensure good reconstruction performance in most use cases (without using too much memory).