Cisco Switching/Routing :: 650x How To Improve Link Failure

I have implemented in my company network recommended by cisco scheme with MEC etherchannel from some segment distribution layer (VSS) to core switches (legacy 650x) exactly as on picture: Core switches  have rather old  IOS (Version 12.2(33)SRB4 and i have reasons not to upgrade it now) which dont support many functions - like BFD over Etherchannels and Enhanced PAgP. And we run a OSPF/MP-BGP/MPLS bunch over those links with OSPF reference bandwidth configured to such values that failure of one link in etherchannel will not lead to link cost changes (we dont have a lot of traffic for now on those links and 1 link can handle it easily) and consequently will not trigger changes to RIB and BGP topology table.
 
Problem is that i am afraid that in case of one of link of portchannel will fail (but without port down event - unidirectional link or smth else) it will take rather long time to converge - and im talking about L2, not L3 protocols. Cos i cant implement BFD and ePaGP over this link - best i can do that improve 30 seconds PAgP hello timer to 1 second. But at least 3 seconds of possible traffic blackholed - is rather big problem.First time i considered 2 possible solutions:

1) BFD - but even i can use it, it will not improve much recovery time cos 750 ms for BFD over etherchannel not much better that 1 second PaGP fast timer.

2) ePAgP - as i understand this protocol have enhancements to discover failed link faster and improve recovery time, but all information i can find about it - is how to use it for dual active detection and none about timers improvements. So i cant be sure that i should upgrade IOS on Core switches and it will improve my failure detection time. ( if it possible adduce some links to description of this protocol enhancements in comparison with PAGP)
 
For now im already thinking about using four L3 links (instead of 2 portchannels) and BFD over those links with 50 ms hello timer. But in case of link failure reconvergence of OSPF will take some time, and BGP will react to this no immediately, so i am afraid that even using aggressive timers for routing processes and hellos i will not win much time.