The number of blocks nodes received averages out to be approximately the number of blocks these nodes send. "Number of takeoffs approximately equals number of landings." If you have a well connected node with lots of bandwidth then it's possible you will send out more data than you receive, but that is unlikely to happen if you have limited bandwidth.  Thus the network average remains 1 to 1 (except for new nodes).

If a new node starts up then it will have to receive each block once.  If an incompetently run node keeps crashing and losing the entire block database and has no backup then this will happen multiple times.   This part of the problem can easily be fixed by nodes with limited upstream bandwidth deprioritizing transmission of older blocks during time of congestion. (One of many possible network optimizations that can and will appear should they be needed.)