I would like to put some words here, maybe they will help.
So when talking about information systems, one tries to develop them and stabilize them by locking the combination of hardware, OS, databases, and applications.
Hardware choice is related mainly to finances and the organization’s capabilities - do you have your own server infrastructure, and people to maintain it, and pay for it, or you can outsource this to the cloud provider. Also, security considerations are present - do you need/like to keep your data within your organization, or you can entrust it to a 3rd party (cloud provider), encrypted or no it’s still 3rd party…
Database choice? Databases stay for 10-15 years once selected.
Operating system - usually 5 to 7 years before being too old to update - atm think about Ubunto 20.04 LTS , before the new LTS version, which will bring different libraries and the software run on it needs to be probably updated/recompiled/adapted.
Finally, the software technologies you choose to implement your applications with.
Haulmont says CUBA 7.2 will be supported for 5 years, for 5 more, pay.
5 years is the average frontend app lifecycle duration.
From that standpoint, I think you should be fine to continue and use CUBA 7.3 as long as it goes, 5 years. You are at 40%, what are you going to do, wait for 6 months until JMix is out and stable and then continue development? I don’t think you can wait, or should throw away what you have. Its standard mix of technologies, you even have source code - bonus is you can use CUBA generators and scaffolding to not waste time on tedious things, and also people that don’t know Java well can work with it. Add-on licenses are enabling unlimited distribution.
Currently, you are using what Haulmont now calls “Backoffice UI”, which is Java + Vadiin.
They say it will work as-is in JMix, plus some more components, but, if you extended these screens to make custom, it will not without a rewrite. Shouldn’t be too hard, especially if you separate business logic and user interface code.
“Scaleable” is very important again with the first item, hardware - nowadays cloud providers will charge you by CPU usage, so you want software technologies with better performance, to pay less.
Also, such technologies (microservices, REST calls) are easier applied if your application needs to scale, for example, you will be adding additional Kubernetes containers as needed.
With JMix, now you have that option to choose if you need it. CUBA is mature as is now, and nothing should force you yet to switch to JMix, just because JMix exists.
Haulmont has a lot of software implementations around, using CUBA, and too many important customers, to whom they are responsible, just to throw it away.