If you want to find unusual problems with older Apple hardware for the sake of reusability, take a look at the firmware from around that time. It was just weird enough to make it similar to standards but often too difficult to support a decade later. In a decade, people are going to say the same about the not-quite NVMe SSD interfaces in current Macs.
Anyhow, I doubt they'd have had time to make a solid 64bit x86_64 SDK even if they had delayed the hardware to late 2006. They had never stopped maintaining the x86 port of OS X (aka Rhapsody, Open Step, NeXT Step), just stopped shipping it as a product, because their strategy was and still is to package their OS in hardware boxes. They started shipping x86 boxes to developers in the form of the x86 development box, which was a (32bit) Pentium 4 shoehorned into a Power Mac G5 case. Its OS is also what leaked to the early x86 hackintosh scene; before Apple shipped their own x86 Macs. I ran it on an custom Athlon64 box and a ThinkPad T42.
Since they already had the x86 port, it was an easy switch at a time when all x86 hardware from Intel was still 32bit. If they knew Intel would adopt x86_64 from AMD, one option could've been to ship with AMD CPU's on the first models, but I'm not sure how well that'd have turned out logistically or whether it'd have made a huge difference.
[1] http://lowendmac.com/2006/core-duo-macs/