In my (admittedly limited) experience, when scanning "assorted" documents like you plan to do, Fujitsu's $500 ScanSnap personal scanners aren't any more prone to jams than their $20,000 production scanners. If anything, they're somewhat less prone to mangling documents, simply because they run more slowly.
With that said, if a few years of tax returns and legal documents will truly take months to scan, you have a massive volume of paper, and you should probably look into outsourcing. Using a Fujitsu ScanSnap, I scanned five years of receipts, tax documents, and correspondence for both myself and my consulting business in a couple days. At the time, I was the IT director for an authorized Fujitsu reseller, so I had half a dozen different higher-end scanner models available for personal use at essentially no cost, but there was no compelling reason to use any of them.
The key to avoiding jams and double-feeds is simply a bit of document prep and paying attention to what you're doing. For instance, when a stack of papers is tri-folded into an envelope and the sheet feeder path is perpendicular to the folds, you will have problems if you expect to be able to absent-mindedly drop the (somewhat folded) stack into the feeder and walk away.
More generally, the key to productive scanning is to come up with a routine that, while you are scanning, is as uniform as possible. You absolutely don't want to be in a position where you have to adjust settings for each document, so, for instance, you should scan everything in duplex, enabling the scanner's blank-page removal if desired, and you should have an "exception pile" where you place any documents that require special handling (e.g., enabling color), so you don't have to interrupt your workflow to toggle settings. Here I'm assuming that the vast majority of your documents will use one "main" setting, while a small fraction will require "custom" settings; if this is not true in your case, adjust the process to suit.
You also don't want toÂbottleneck the process with synchronous processing, so if "in-line" OCR and/or image processing is materially slowing down scanning, you're probably better off disabling it and doing these tasks by other means (in batch when you're done scanning, via tools that watch a folder and process documents as they arrive, etc.).
Finally, unless you have some form of robust automatic document classification, I've found it's far more efficient to separate "scanning" from "organization" — first, I scan everything into an "unfiled" folder with sequential, generic filenames, then once I've finished scanning, I use a combination of the OS X Finder's "quick look" feature and ad-hoc, keystroke-activated scripts to label and sort the documents. If you can enlist an extra pair of hands, you could clearly improve throughput by running an "assembly line" where one person scans and the other organizes.