What if the blocksize limit were chosen by each node individually?
Everyone would set the minimum = send and the maximum =receive.
So for example I would send minimum 2 mb blocks (if full), but accept maximum 6 mb blocks.
Would it work? That would make it optional for those with shitty internet to put a cap, but those that have good internet would have bigger cap (but not infinite to defend against DDOS)
changing the block limit individually, to reject blocks.. is a stupid thing to do purely for slow connections.. it basically renders the node no longer part of the network
but here is some advice.
limit the number of node CONNECTIONS and you get the desire you want.
EG someone with fast internet can have 50-100 nodes connected. (it used to be termed supernode)
EG someone with slow internet can have 1-6 nodes connected.
the point of the blocksize consensus is that everyone agrees to a certain level.. you may be asking why..
to ensure everyone accepts the same block data. and thus no rejects/orphans due to relays. (there are other ways to orphan but pools can mitigate those other risks as explained before)
so getting to your previous point. if everyone was at 2mb.. they can ALL accept any blocks of 200byte-2000000byte with no issue. meaning a 0.5mb block is accepted, a 1.001mb block is accepted a 1.9mb block is accepted
so by having "block consensus". people only then need to mess around with how many nodes they relay to and from. which can be independent. and a post on another page of this topic pointed out the 37mb/10 minute upload ability of a 0.5mbit internet connection (baseline) shows that even slow internet connections can relay to a few nodes..
by the way its worth pointing this out because even segwit goes upto 4mb(buffer) of REALDATA
*.. and if you are thinking of running pruned/no-witness mode, you are no longer a full node so might aswell just run multibit or electrum or other lite clients
now preparing for your rebuttal that not everyone would accept 2mb..
this whole year has been about the 2mb buffer wont be ACTIVATED until consensus is reached. and so it wont make a difference before consensus because noone will make blocks over 1mb before consensus.. and obviously consensus is reached if it got activated which means everyone would be accepting the new buffer
*segwit keeps "base size at 1000000 meaning traditional transactions are at the same 1mb limit.. the extra 3mb is only for segwit witnesses(signatures)
which with a 1in-2out standard tx segwit doesnt offer much more capacity allowance at all, but does allow the capacity trick to be noticeable more if everyone was using multisigs or multiple inputs.. dont confuse segwits 4mb buffer as a "blocklimit" buffer of 4mb for traditional transactions.