Since Keepalived and Cisco/Force10 both use VRRP, and use an ID Range of 0-255:
1) Would using the same number in the same vlan cause a issues.
2) Would using a different number in the same vlan cause a issues.
3) Would using the same group ID on another vlan cause issues.
I am thinking of reserving the last 55 group IDs for Keepalived and the first 200 for L3 Switches.
A VRID needs to be unique on each broadcast domain (VLAN) so as its used to uniquely identify a Virtual router and used to generate the MAC address for that Virtual router so to answer your specific questions
1) Yes, If KeepAlived and your Cisco routers are using the same VRRP ID for a different set of IPs in the same VLAN you will have issues as they will have mess with each others master Elections and ARP addresses.
2) No, Different Number in the same VLAN is fine. this is the intended way to have multiple Virtual routers running in the same VLAN.
3) No, the design of the protocol is that VRRP message should never be routed so they will never be forward across broadcast domain(VLAN) boundaries.
The end result is that you can have 255 Virtual routers in the same VLAN as long as you ensure that each IP is only bound to one virtual router and each virtual router has the a unique VRID. I think your method of dividing the ID range is sound, Its big advantage is that you don't need to a third seperate system to track VRID usage. As long as keepalived sticks inside its range and ensures each keepalived cluster in the same VLAN is using a unique ID then it doesn't need to be aware of the VRIDs used by your routers(and vice verca) the draw back to this method is that your setting low thresholds on each service. (ie keepalived can only use 55 VRIDs even though routers may only be using 1 of its 200 ID range. My guess though is that draw back isn't going to be an issue, If you are more than 10 VRRP groups in a single broadcast domain(VLAN/SUBNET) then the subnet is too big and probally needs to be carved up into smaller chunks and your likely to hit issues with too much broadcast traffic beofre you run out of VRIDs.
Generally keeping the VRID's unique everywhere should make any packet dumps, debugging or logging easier to decipher.