Hi,
If MIPs are configured on different interfaces the correct MIP is always selected automatically. This is the MIP configured on the egress interface. Everything depends on routing. The preffered ISP link is the one which has a lower metric on the default route and the MIP configured on its interface is used. To initiate a routing failover to the secondary ISP and switch to another MIP, if the primary ISP has failed, you should configure IP tracking on the first interface. It's better to ping a remote but good reachable IP for this. Sure, you can also monitor the CPE router IP but it is not always a good solution. The ISP may have routing problems in the backbone while the CPE router IP stays pingable.
IP tracking puts the interface in the status Down and deactivates its routes if tracking has failed. The default route with the higher metric takes over the traffic and uses the second MIP.
You can also use a primary-backup interface method. The difference is that the backup interface is always down under the normal conditions.
You will not be able to implement a good load balancing solution without an additional device (because of NAT). Theoretically it's possible to configure load balancing on a per session basis with Equal Cost Multipath routing (ECMP, up to 4 routes with the same metric) but any application, that establishes multiple sessions, will fail.
What is possible is load sharing. Source based and/or policy based routing can be used to statically distribute traffic over the both links.