I prefer that this is in a separate merge request (during past month it was working in !383 (merged)), because this is in fact about migrating python to high precision. There might be more instances of such problems found in the future. I have written about this in a compatibility and debugging sections in the docs, so that they might be fixed faster.
In this specific case it turns out that numpy linspace function does not understand mpmath. Which is surprising, because (so far) everywhere else numpy works with mpmath.