anoncvs is failing with ENOSPC type errors from the remote server.

I am running recent cvs code.  I am losing on reintegration of
smallish (3-8K loses, 1K is ok) files.
The symptoms are that a bunch of packets get sent to the server and
get strung out in time by the modem.  I don't see any loss.  The same
behavior on the ethernet results in a clean reintegration.

In connected mode, writing an 8k file results in a clean store.

I am running with reduced parameters for rpc2

    venus -rpcdebug 1 -timeout 1000 -ws 2 -sa 1 -ap 1

and the trace of the store matches this ok.

In disconnected mode, I don't see the same pattern - the client sends
more packets without waiting for acks, and the reintegrate/store
fails, and the client goes disconnected.

So there seem to be two things going on:

  rpc2 params are not taking effect on the backfetch (are they
  controlled by the server?)

  there are some sort of timeout problems or something else that is
  causing reintegation backfetches to fail over 28.8 modems


Any clues?  This is making coda somewhat unusable for me.