I get the feeling that there are two separate issues going on:
1.) ISP's may be throttling back some users as they think that the Final Fantasy XIV client is Peer To Peer (P2P) sharing. Some ISP's do this to stop gap large sharing clients (BitTorrent) causing latency in their networks.
2.) There are massive latency spikes on the server end causing "rubber banding" and delays when executing actions and movement.
If the ISP is throttling the actual client, one wouldn't see low ping rates followed by some extremely high latency followed by a drop to normal again. Throttling would be a low connection all the time with little to no jumping around (see my post here for reference
http://forum.square-enix.com/ffxiv/t...ottling/page76 )
"Rubber Banding" is a symptom of packet loss and poor end point connection. When the server (Square) has to send the client (Player) packets there has to be an acknowledgement of the data on the client end and a confirmation of the data being received on the server end (Think MD5 checksum). If the packets are incomplete, the client can't confirm with the server it received all of the data so it tries again. In the meantime, all data stops parsing (the flow of data is interrupted). Once the packets are confirmed, the server and client then have to double down to send data and present them visually while syncing their clocks together (the speed up and all of the logs spilling in at once).
Hopefully Square can set up another sticky for the second issue (Easily duplicated. I've even set up a template for users if need be) so we can store aggregate values there and not clutter up an existing issue.
In the end, it may not be necessary as the "rubber banding" issue seems to stem from the MPLS (Ormuco) and rack (Tata) end. Until the MPLS circuits are upgraded to accommodate more bandwidth and the server ports are expanded and prioritized, latency and route paths will be impacted.