Interesting and useful post. That being said, I think it ignores some valid solutions. For example, REAPER and some others offer the option to bridge specific plugins. This is often sufficient, since users generally just want to play with some specific cool-sounding but crashy plugins.

Then there's also the more technically difficult option of running the entire audio engine (including plugins) in one separate resilient process: if it crashes, it can be relaunched and its state restored by the main process. I believe Bitwig can do this. Cool but I wouldn't want to be the one implementing it!

In the Ardour world, the "bridge specific plugins" approach is accomplished by using JACK. There are several clients that can be used to run a single plugin (for at least a couple of plugin APIs). The level of integration drops noticeably, and it does mean that you have to use JACK which for some people seems to be a burden. But it also leverages the generality of JACK's inter-application audio/MIDI to avoid complexity in the host itself.

As I said in the article, we are averse to complex engineering solutions to the problem of plugins that crash.

Wow! Thank you for sharing. I really appreciate the work you do, by the way!