I do have a database
I still miss some things a spreadsheet can do.
I'm right now torn between different modes to analyze data.
Right now I'm R-centric.
But in R you do everything in the command line. all changes are logged, which is good. Spreadsheets are a big no-no in this community.
The rdbms community has different ideas about how to manage data.
Basically you get what you want a query at a time.
And of course, Joe six-pack does everything with excel.
I wonder if I'm not missing much of the attractive of a visual way to handle data. Color-code cells according to contents, zoom in and out with an slider, ... that kind of feeling. Very exploratory. Compare that to a command line where you do head(data)
Since emeditor and R prove that you can do lots of operations without placing things into memory (or screen) I see no reason why one cannot have a streaming spreadsheet that loads the parts you need on the screen really fast.
Doing operations by selecting block would generate the equivalent SELECT for the db... if you need one. There might be very fast alternatives.
I think there are many paradigms for data handling that are not that well explored. For example, comparing OO, RDBMS and RDF (linked data) is eye-opening. This is beyond the point I'm bringing here about a streaming viewer (btw GUIs for RDBMS do suck at this; it's not that hard!)... but still. Let me know if you want more info or if I'm not making sense at all. I have a paper in mind on Object-relational mappings that is very good, if a bit technical.