Rethink of object storage for pile-up studies
Object storage over multiple events
As of !94 (merged), all messengers keep a shared pointer of the messages which they have sent, until the end of the event. If one would do studies where data from previous events is required, this HAS to be copied locally. Aside from the efficiency of this, an important aspect is that the MC history for these copied data is lost, unless a corresponding copy of all historical objects is made and linked at the same time.
Note that it is not possible to fix this by re-implementing the NO_RESET delegate flag, as there is no option to subsequently remove the data and it will be held forever (the module which holds the data is the module which produced it, not the module which will use it and decide when it is no longer needed).
Proposal
Some more thought...