Closed
Description
See bottom half of this comment as well as this later comment.
tl;dr something we're doing in Runner - I'm guessing some of the ioctl junk but not really sure - is messing with shell backgrounding such that the process has to be foregrounded again to continue execution.
It only occurs when a real pty is attached; headless processes (like the kind that used to break before #425's main issue was fixed) don't exhibit it. Which is why I'm guessing it's to do with the "passes isatty/has_fileno tests" code branches.
Issue started around 0.12.x in case that helps us narrow it down.