in my experience sequence diagrams tend to betray the challenges of distributed computing.

Do you mind explain it a bit more? I use it mostly to document the steps between systems, I don't see how distributed computing changes that.

there's no guarantee that 2 requests kicked off at the same time will arrive or complete at the same time. When time is an axis (often y axis), typically we see parallel lines in sequence diagrams, when in reality they will often cross the latter request will actually arrive/finish before the earlier request.

Additionally, any reasonable system will have timeout + retry, but it's not possible to differ btwn timeout and inordinately delayed. Meaning your requests maybe replayed many times, and the system frequently doesn't handle that well