I see it more like if a project is extremely difficult to implement (has a high rate of failure at the implementation level) then the most critical phase where something can go wrong is right now (the code writing phase). What I mean is the reason TML is taking so long is because the project is so ambitious and there can be zero bugs in something as mission critical as the core of TML. So you are right it's about quality/reliability and maintainability. 

It is also the fact that the theory behind what Ohad is doing is so complex that at this time no one can help him. So it's still at that immature stage where he's still figuring out how to code it himself and no one really knows his vision better than he does at this time. When it gets to a point where the most important components are coded then you will see other programmers begin to chip in and the speed of development accelerate. 

This is not like Ethereum with Solidity. Some people believe Ethereum was a rushed design which  was hacked together for speed of implementation. A deliberate approach also could have been taken with EOS which also had major flaws at launch. If you're going for provably secure, provably correct, decidable, etc, then yes it's now more complex.

A compiler alone could be easy. In fact if I were going to do it I honestly would have hacked something together from other projects which have compilers. But Ohad wants to do it all from scratch to meet his specific and precise criteria. Because nothing similar exists this means he has to research and learn while he's coding which seems to be why it takes a long time. He has to determine whether to go with bwd chaining or fwd chaining, or why pfp is the only  suitable logic and what the consequences to the design will be because of this.