rudder agent update doesn't always update policy
We use several flag files to copy policy (promises, ncf, tools) from the Rudder server to Rudder nodes. The way these are used is:
1) Copy flag file
2) Check if copy changed flag file
3) If there was a change, start copy of main contents
However, if step 3 gets interrupted, or fails for any reason (not enough space on device, network issue, user killed process, etc), the next "update" will stop at step 2 (there will be no change to the flag file).
"rudder agent update" should be able to get around this, using a "force/-f" flag.