The thing is, you cannot submit to a "lower-tier journal"; once you get accepted to one journal, no other journal will allow you to resubmit the same results (even with readability improvements).
The way I've seen things done in CS these days is not so bad, though:
Iteration 1: When you are done, you submit to arXiv and simultaneously to a conference. Conferences usually have as strict page limit, so many things get swept into the appendix.
Iteration 2: You get review comments from the conference; depending on the rigor of the reviewers, these may or may not be ideas how to improve the text. You submit the camera-ready version of the conference paper.
Iteration 3: If the paper was more extensive, you start preparing a journal version, where you can move things back from the appendix make the text cohesive again. You probably also have feedback from the conference participants, which is often extremely useful.
Iteration 4: You submit to a journal, you get a (usually quite thorough) reviews and you integrate them to the final version of the paper.
---
Bottom line: I'm saying it's not so bad (compared to just a single iteration) but I hate this system. First, it does not allow community contributions from outsiders (such as PhD students who read your paper and notice something small that can be improved) and second, after Iteration 4 it is nigh-impossible to make small improvements to the paper unless it was spectacularly wrong.