It's great you're exploring alternate gantry designs and drive systems.
It sounds like you've only done a little research so far. There are some techniques gain expertise, and this is essential to pull off a machine of this scope. I'd suggest using Q&A to resolve the harder or unclear issues, and verify what you've learned already.
On R&P, for example, try a search like this:
https://www.google.com/search?q=rack...n+vs+ballscrew
Here's shaft whirling/whip in action -- it's a resonant phenomena, so worst at certain frequencies.
If you recall from Mechanics of Materials, beam deflection is proportional to Length^3 or Length^4. This is why it's difficult to make a ballscrew work at long lengths, and also unsupported round rod.
On linear guide/shaft/rail, for example:
https://www.google.com/search?q=linear+shaft+vs+rail
And see the archives for thousands of posts why unsupported rail is flexy.
When starting in a field, one of the hardest parts is knowing what keywords to choose, and of course google results are highly dependent on keywords. A good strategy is to search on what you know, then read mainly to gather better keywords, then search on those new keywords, and repeat. The best designers are sponges for good information.
As Gerry noted, repeatability is necessarily always better than accuracy.
A practical problem with a 3" thick solid gantry is it can't even manage its own weight. Let's say it has 0.005" bow (fore-aft) in the installed orientation, but we didn't know that and haven't measured it. Now lay it flat. It's self weight will pull it flat to the mill table. We put feeler gauges under it and detect nothing, so we think it's sitting pretty. Then we skim the rail mounting surfaces. Then we stand it upright again: 0.005" bow returns. You can run the beam deflection equations to verify. A big tube will be much more stable because it has much higher I and J, and much less weight/foot. Sure, we could indicate the solid bar before laying it flat, then shim it, but why work so hard when there are better designs.
Closely study the industrial routers: TechnoCNC, Multicam, Onsrud, Thermowood, etc. What do they do for gantry design, linear guides, drive systems, etc? Try to find a ballscrew on an axis longer than 5' -- that is very rare, and if so, big diameter ($$) for the reasons mentioned. As a rule, when one is not an expert in a new field, it's wise to assume that the standard designs are standard for good reasons, and that non-standard designs will probably have problems.
There are some general guides out there on DIY CNC, e.g.:
Official Home Page of the Roman Black
Mostly correct:
Homemade CNC Router The Builder's Guide (FREE!)