I'm concerned that xtask ci will lead to less optimal practices as presented:

• Some CI steps you want to run under every environment (test) while others you only run once for performance (rustfmt, clippy) but this instead couples the CI tasks together
• Some CI steps you only want to run under special environments (RUSTFLAGS=-D warn cargo check for consistent warnings) while others you need to run in every environment
• Care is needed so that both the users and the CI do not think the CI is hung.
• Can't leverage CI UX features to call out what step in the process failed ("CI failed" vs "clippy failed")

I say this unsure whether to drop this xtask or to find ways to mitigate these concerns.