I'm been trying to setup serial IO for my brother for months (I've got it setup for myself) and I thought I had all the kinks worked out, but ran into this same issue where the test call worked but the daily call didn't. Then when we switched back to use the modem and we ran into daily call problems as well. After a bunch of packet sniffing, it seems that the service is getting to the point of sending data in raw TCP chunks, but it's taking the series one tivo a long time to send the ACK that it received the individual chunks (don't know if this is normal behavior for the series one or not). The service will send the next segment while waiting for the previous ACK, then when it doesn't get the ACK quickly enough the service is retransmitting the first chunk. It does this send two chunck, times out for the first ACK, retransmit first chunk, get the original chunks ACK sequence then continues. I ran into problems because I had set my pppd timeout to be 30 seconds and this process got to the point where it was taking longer than 30 seconds to catch up to the data chunks. Unfortunately, at the same time that I tried lengthening the timeout (to 180 seconds) I had just previously been trying it with the the speed bumped down to 57600 (from 115200). With the longer timeout, I'm getting much more data (it's at Ack=38580 after 19 minutes where before it would timeout after Ack=10180 at around 5 minutes). Also of note, these segments are small (only 1420 bytes) each and the time between ACKs slows from 6s,12s,24s,48s,96s,120s,120s,120s... It starts at 6 seconds, then doubles until it tops out at 120 seconds between ACKs - very odd. I'll let this work for a while longer and send an update if the elongated timeout is a viable work-around for other serial IOers (though this is very painful to have to deal with all the retransmits that must be really slowing things down).