Take the opportunity to get your address planning done and implementing an adaptive scheme.
Check the activity levels of the ELANs, look at where you actually need the subnet/VLANs, and make your adjustments such that you can do efficient summarization with your routing protocols and gain the best advantages of redundancy schemes.
Evaluate your cabling plant and redo/supplement/ update any area where aggregated LANs are likely to pass.
Anywhere you are using ATM QOS will (probably) need to have some flavor of Ethernet QOS. If you need to carry the QOS across the topological boundries, check into whether MPLS will work for you and figure out how you should implement it (and how well it will phase in).
You mat also want to look at what will be your first area for GigE implementation. You really can't achieve ultimate throughput with GigE without using Jumbo Frames and other adjustments that are not (usually) suitable for older legacy equipment. You may need to keep the GigE throttled back until a full span has been updated ... but you should know that in advance, and make it all part of the implementation plan.
Be prepared for some extra latencies when crossing topological boundries. The extra time in restructuring the cells-->frames/packets-->cells (and VV) may "slow down" some critical traffic.
It really boils down to making it as easy or hard as you want ... as long as you know the trade offs down the road.
You might want to do a final baseline before rolling out the new stuff, so you have some basis for comparison once the new stuff is in and lit up.
.02
Scott
