We have the same problem on our 1500Ds FGT with 5.6.6 version of OS. It is configmed as a bug:
Bug is fixed in FortiOS 6.0.3 GA and interim build 5.6.7
We have to restart every 3-5 days the WAD process, sometimes the worker sometimes other process.
When the 5.6.7 will be out the support could not tell as they did not have this info when I asked.
We have a customer with 1000Ds in cluster and had the same issue on 5.6.4 and we updated it to 5.6.5 almost two months ago. They have did not experienced any issue so far. That is the reason we are not recommended to upgrade to 5.6.6 as it is confirmed having a bug.