Try switching back to async pipes after next update of Linux distribution
While switching to blocking pipes together with block_in_place
resolved #200 (closed) for us, this situation is less than satisfactory as it shows that the protocol itself is not problematic and pipes should allow us efficient asynchronuous communication with multiple external harvesters instead of using blocking I/O inside an asynchronous application as a workaround.
Hence after the next update of the underyling Linux distribution we use (currently Ubuntu 22.04 LTS), we should try to change the external harvester interface back to use asynchronous I/O to see whether the change in base system resolved the lost write readiness notifications.
Edited by Adam Reichold