The backfill process goes through a separate "pipeline" outside of the normal sync, so it won't compete with your regular sync cycle. However, when there's an update to a field value in SFDC, that creates a new SysModStamp on the record, indicating that it's been updated and should resync to Marketo again.
When a lead converts to a contact in SFDC, the field values from the lead record just transfer right over to the contact record. So with the backfill pushing data into the new fields, the data going in will get copied right over to the contact record once it's converted (assuming you've recreated the fields on both the lead and contact objects).
The main goal of scheduling the times for when to recreate the fields is to avoid a bottleneck of having multiple customers kicking it off all at the same time. By spreading it out over time, we spread out the workload on the data transfer, so it keeps it all in check. That said, we've had a surprisingly low number of customers actually making the changes so far. So much so that we're actually wide open if you wanted to kick it off early. Big disclaimer for anyone reading this later on -- this applies as of the scheduling right now and I can't guarantee this will be true later on during the deployment. As of now, we can move the schedule forward if you wanted to kick it off now, but we can't push it backwards if you wanted to put it off. With the low turnout so far, we're also expecting a lot of people to rush at the last minute, putting stress on the bandwidth later on. So... short version there is, go right ahead and start it now. 
.png)