I could easily lose all that I have gained. I am really certain that this is highly likely if the modems have only been up for a short while.
However, the long term observation, formed over a number of years, is that in my case if I leave modems on for a really long period that is without problems and so no spontaneous resynchs, then they will tend to resynch higher when a resynch is forced. Obviously there is a ceiling, otherwise by that argument sync rates would be pushed up and up for ever. I think in part this is the intelligent god overriding statistics, because if sync rate is low then you choose to wait a good length of time and then resync and if the rate is high then you do not resynch, so perhaps it is that choice intervention process that biases things and it is as much about the properties of the god of modems (me) than about DSL.
I wonder if this is totally not true in the case of other lines, even if done with the same intelligent intervention strategy.
The other candidate explanation is the ‘change in conditions’ one mentioned earlier, and for that it would have to be the case that there is long-term drift away from the initial state, so the original plan becomes less appropriate (and if fact shifts to something beyond what bitswap can adapt to), rather than cyclic variations.