E.g., when a small-sized block arrives, delay it for 3-5 seconds, and only rebroadcast it when no new valid, larger block is received during this period. Would it encourage mining pools to increase their blocksize, or does this behaviour have any bad effect on the network?
I am pretty sure all major miners (pools and solo miners) ensure they are well connected to other miners. It doesn't really matter if non-mining nodes delay blocks a few seconds. No miner is going to delay building off a block which extends the longest chain as it would only negatively affect their revenue.