Thanks Mark. Will do!
The VBUK~0 index was always there - perhaps my included analysis was not the latest, or perhaps I was rebuilding the index at the time (improbable but possible).
These are two different DB's on two separate servers "EQ1" and "EP1" of our QAS and PRD SAP systems respectively - though as I mentioned the '210' (slow) one was copied recently from the '300'.
Comparing line by line, I am noting some differences in the 'cluster' ratio and other values which obviously affect the cost determination. I did rebuild the index early on and had run update index statistics, but I wasn't methodical about it at the time.
I will follow up with your suggestions in that area and continue to compare the analyses to see if I can pick out a relevant factor.
And thanks so much for taking the time to go through all this stuff :-)