Hello John,

MPLS Te has its own signaling plane based on RSVP-TE and TE extensions of link state protocols.

So it doesn't use or interact with IP SLA.

to be noted that if you are in an MPLS scenario you don't need to redistribute BGP into OSPF.

the mpls forwarding for recursion follows the LSP path to the BGP next-hop (loopback or remote PE).

so if you multiple paths to BGP next-hop CEF based per destination load balancing using exor of IP SA, IP DA, and hash should already happens.

if you redistribute the BGP networks in IGP LDP will create additional LSPs for these IP subnets but the load balancing should not improve.

if you have two paths to BGP next hop you will have two paths to VOIP subnet

but using per packet load balancing has negative effects on jitter (delay variation)

Hope to help

Giuseppe