As an interesting note, we don't node_format(pidfile) when we detach vs when we remain attached. But pidfile names seem to work as expected anyway so presumably that formatting is happening somewhere. A %n in the pidfile argument gets handled by the multi runner, but then if I add a %I, it gets mysteriously handled somewhere on the way to being used to drop the actual pidfile once the multiple workers get started. Probably not an issue.

can you check why pypy tests are failing?

That failure is unrelated.
@auvipy @maybe-sybr Can you please open an issue so we'll eventually investigate this?

#6489 @thedrow . I also just re-ran the travis actions.