How do you handle distributed transactions?
You don't particularly need distributed transactions if you design it right.
Generally try avoid them if you can. They have massive performance overheads.
The problem is that more often than not, you don't know the requirements beforehand.
I'm trying to come up with examples of transactions we have that span multiple services but can't come up with any. We do group tightly related functionality in the same service if we can. So if they need to be updated strictly transactionally they can usually be in the same DB-transaction.
Any particular example you'd like my thoughts on?
I work in a microservice codebase with the realization that it might have been premature to what we needed but it does roughly live up to our aspirations. Products sharing infrastructure and common functionality (accounts, file storage/management, notifications, email) and solid boundaries for easier outsourcing, support for using different languages in different parts if necessary.
I've found gains, I've found costs. The article is right in that microservices are not a silver bullet, I think few people really believe they are.