I fall somewhere between 1 and 2. I would use many of the items ahead of cross DB search more frequently, but I agree multi-DB search is important, so I'll support your survey with a +1.
Part of Kinook's calculation is certainly an estimate of how much work each roadmap item will take to implement. Some of the items ahead of multi-DB search will probably require a lot less time even if a few of them are less critical.
A bit off topic, but I'm concerned about a different roadmap item--desktop search--being the kind of feature that could bloat UR beyond it's core purpose. I'm glad the roadmap says "investigate". We already have free desktop search tools like Windows Search, Google Search and Copenic.
Hopefully I'm not misunderstanding the intention, but IMO, indexing file names inside linked folders is about as far as UR should go down the desktop search road and even that behavior should be off by default. I'm concerned about overhead, convoluted search results and size of the index.
On the other hand, Multi-DB search should be considered core functionality IMO.