Big blocks are not synced correctly

[ahaensler]

While syncing a range of large blocks (200kB per block), the following error keeps happening

2020-06-27 07:43:32.567 tokio-runtime-worker DEBUG libp2p_swarm  Connection Connected { endpoint: Dialer { address: "/dns4/netman-blockchain-interface/tcp/30333" }, info: PeerId("QmUnBwbLUgyd18NS9Q9rrDExDTBMGk33Nd8VjUFUDmLYTS") } closed: Handler(Handler(A(A(B(Upgrade(Apply(TooLarge { requested: 18159734, max: 16777216 })))))))

and peer connections are closed with Reserved peer QmUnBwbLUgyd18NS9Q9rrDExDTBMGk33Nd8VjUFUDmLYTS disconnected. The node is stuck and cannot download blocks.

The 16MB limit is set here https://github.com/paritytech/substrate/blob/master/client/network/src/block_requests.rs#L141
Can we increase the limits? Or expose these constants so they can be set by a service builder?
It would be nice to log the failure as error instead of debug, there is no error log currently.

[bkchr]

CC @tomaka

[tomaka]

Should normally be fixed by #6398
The sender will now send fewer blocks in order to not go over the limit.

Please reopen if that's not solving the problem.