I've never found a way to automate their removal. I examine every glyph of every interpolation and remove them manually. I probably spent a 3 or 4 hours last month removing those things from a large interpolated family. I asked about this a few years ago (I forget which forum) and nobody had a solution.
It's odd how some times it doesn't occur. Sometimes exactly the same glyph will have those errors on different diagonal lines.
Similarly, you should keep an eye out for any glyphs with identical components. For me that's usually %‰…“”„"«»÷. The identical components will get swapped, seemingly at random. It doesn't matter if the components are ordered correctly. You can even interpolate the same fonts multiple times and get different results...sometimes. Changing the order of the components makes no difference to the outcome. If you deliberately put them in the wrong order, the results are the same. Strangely, the problem occurs for all glyphs with identical components or none. It's never just one case...which is good because it makes the problem easier to spot.
If I'm making something new I'll use multiple masters but often I need to interpolate older fonts. If I only need to interpolate a few glyphs, it takes less time to blend and clean up than to set up multiple masters. For example, if I need to add new some currency symbols to an existing typeface, I blend, clean up, blend, repeat.
"Blend" can work with otherwise less-compatible outlines, which is handy for some things—although often just to show/explain/educate, rather than for production work, in my usage.
Comments
It's odd how some times it doesn't occur. Sometimes exactly the same glyph will have those errors on different diagonal lines.
Similarly, you should keep an eye out for any glyphs with identical components. For me that's usually %‰…“”„"«»÷. The identical components will get swapped, seemingly at random. It doesn't matter if the components are ordered correctly. You can even interpolate the same fonts multiple times and get different results...sometimes. Changing the order of the components makes no difference to the outcome. If you deliberately put them in the wrong order, the results are the same. Strangely, the problem occurs for all glyphs with identical components or none. It's never just one case...which is good because it makes the problem easier to spot.
"Blend" can work with otherwise less-compatible outlines, which is handy for some things—although often just to show/explain/educate, rather than for production work, in my usage.