What factors prevent companies from shifting to distributed network architecture?

A number of good possibilities has already been called out. Clearly, investments made into creating architecture you need to shift from (hierarchical) would create some friction, as would a potential lack of experience with the target (distributed) architecture. And of course ultimately, the network architecture is there to support workloads and access methods, so there has to be alignment. 

A couple of reasons:
1. Management of deployments across a variety of clouds and on-prem infra.
2. Latency added due. to communication between data centers.
3. Ability to diagnose and resolve issues in a live application.
4. Architectural shift from the monolith to microservices (not everyone needs it).

Couple of observations that I have noticed:
- Most of the start ups start on small scale and at that time it is not needed. They chase hyper growth in business but no growth in the tech stack
- This is mostly due to business leaders unknowingly not understanding the important of technology when the scale of the business is beyond MS excels and simple automations. 
- By the time the ship has started sinking due to holes, the cost in terms of time and money further pushes the prioritisation
- There is also somehow a lot of friction in budget allocation for these initiatives as they are more forward looking initiatives and will not have been immediate impact on quarter or annual p&l. A few business leaders understand it and those who do generally have build successful start ups. 
- Due to above situation the engineering talent within the ecosystem doesn't develop at that maturity and some of times there is lack of intent from engineering leaders to pick these up as it will require them to push their own boundaries.
- Also in the market too the talent for bootstrapping systems is much readily available compared to building matured engineering systems. 

Mostly legacy infrastructure in my experience, and 'server' hardware in particular. A distributed architecture calls for a broader set of hardware that some organisations are failing to raise the capital for. That last point is increasingly difficult when much of the industry is moving to a pay-as-you-go and software-as-a-service model.

Mostly legacy infrastructure in my experience, and 'server' hardware in particular. A distributed architecture calls for a broader set of hardware that some organisations are failing to raise the capital for. That last point is increasingly difficult when much of the industry is moving to a pay-as-you-go and software-as-a-service model.