This save roughly 1 second in removeforblock and then another sec saved for creating a block template for a large 100MB block.

• just some timing is off and we need to wait for items to show up in the mempool.

as frequently as is actually needed. It's a relatively expensive operation, although only taking a couple of microseconds, when multiplied by 500,000 or 1,000,000 it starts to have a negative effect on template creation. Especially given that most transactions don't have parents and therefore this function call rarely needs to get executed.

• When CaculateMemPoolAncestors() is called we don't need to check the mempool parents that are not fully in the mempool. This is really only needed when we do consistency checking.

• Because of all our multithreading we have to properly wait for each chain to catch up to it's intended height before shutting down each node and restarting again.

Lately it seemed that our weekly gitian CI task fails predictably
while building the base docker image. Added the creation log to
the set of artifacts so that we can have an hint of what's going on.

Lately it seemed that our weekly gitian CI task fails predictably
while building the base docker image. Added the creation log to
the set of artifacts so that we can have an hint of what's going on.

• We don't need to see this statement all the time, but only when we have some debugging turned on.