This issue seem to happen with CLAMP operation in SB00 lab.

see rocketchat-policy 6/6/19:

"summarizing .. can reproduce the error, mimicing clamp behavior with back to back deletes and post to api and pap, the error is in the deploy .. adding a delay right before the deploy fixes the issue. it seems a transient issue bounded in the interval between the create guard policy in the api and the 3 policy set deploy in the pap. Based on that it seems related to the data layer, which is the common piece between the 2 components, the idea is that something does not get updated by the time the pap attempts to issue the deploy. Adding a 3 second delay between the create of the guard, and the deploy solves the problem. This is reproducible."

See files attached for script and output reproducing and solving the problem.