As for experimenting, I'd say (and this community can be my witness 🙂 ) that managing your own experiments isn't a great idea. First, you have to maintain the infra (whatever it is, a tool your wrote yourself, or an excel sheet) which isn't fun and consumes time. From what I've heard, it usually takes at least 50% more time than what you initially think. And since there are so many tools out there that do it for free, then the only reason I can imagine of doing it on your own would be if you have VERY unique needs.
I think ClearML has an advantage over MLFlow and the likes as first, the UI is a bit more up to date and modern with nicer looks and better features, but then when you consider the fact ClearML also has data management, and machine management and these goodies, it becomes and easier decision (In my eyes) not to spread yourself.