Router lifecycle events occurring when large numbers of addresses exist causes Agent heap spike

[k-wall]

If a router event occurs (such as a router restart, or the addition of a new router owing plan rules) whilst large numbers of addresses are defined, the Agent's heap memory usage may spike possible leading to an OOM condition.

[k-wall]

One problematic case is router_config#ensure_elements() - here promises are used to cause the deletion of n stale addresses and the addition of m new addresses. In the case, where say a router restarts and the number of addresses large, this leads to an equal number of promises being created. Even though the management requests are throttled further down the chai we still have the overhead of the promise itself. I believe this will explain a spike in memory in some circumstances.

[k-wall]

Reading around on the internet, I'm lead to the conclusion that creating thousands of promises at one time is a bad thing and leads to memory spikes. I plan to throttle the creation of the promises representing creation and deletion of address. At the moment, https://www.npmjs.com/package/p-limit seems to be may favourite - it is quite unobtrusive.