Lock-in is a non argument isn't it?
I liken lock-in to trying to get the last bit of air out of a bouncy castle you hired for your kids' birthday party: you stamp on the bubble of air in an attempt to get it out and it pops up somewhere else.
Virtualisation is held up as a panancea to all of this but, pardon me, where is the openness of the Hypervisor? Sure, you get 'freedom' with your hardware (coughs - anyone tried building a production VM farm with differing hardware and chipsets?) but the lock-in moves to the Hypervisor.
"Go for commodity servers" I hear the cry. Well, all the stuff that sits on top of it (build processes and management tools etc. ) *always* have hooks linked to the hardware. I was there when procurement came in and said "I know we decided on HP 18 months ago and we're about to go live but, my, those Dells are cheap". EEK!
And even if you do get some amazing abstraction to allow complete freedom in the infrastructure layer the problem then moves to the complexity you'll create for yourself in the evaluation / RFP process (cue an army of sourcing advisers queuing up to extract your firm's money)
So - I suggest lock-in is always there (you try and move your core banking application or ERP system at short notice to anything else or change one of the component parts) and it's about recognising that and ensuring you and your providers have a good, open and mutually beneficial working relationship. I think you need to ensure you have the option of change but don't let that drive some religious debate about vendors.