(My answer to this question at Quora.com)
As Amir Aharoni points out, wordprocessing in Hebrew—especially when it involves mixing it with Roman text—is still a largely unresolved minefield, with certain problems “baked in” to the standards for Hebrew display in computerised contexts. Disruption of order—of words, of letters, of Hebrew and Roman text—is one of the most frequent problems. Even if you manage to get it to display properly in one file, once you do anything to it—convert it to another format, cut-and-paste it to another file, or even just insert a punctuation mark, such as a comma—it will likely be disrupted.
Here’s an example of letter order inversion in an email (ironically, from the Standards Institute of Israel, whose Computerization Committee assured me fourteen years ago that ‘all issues [regarding computerized Hebrew] have been resolved’):
Sometimes, the most inexplicable combination of order inversion occurs, as in this example where you have an inversion of order of a pair of words , of clauses either side of a comma , of Hebrew words  text before and after a comma is swapped, and the letter order of just one Hebrew word :
Another common phenomenon is that the word order display may appear right in the composition window, but wrong in the actual display of the text in published form—as in my:
The only practical workaround in my experience, to ensure that the order remains correct is to place the Roman text on a new and separate line, then carry on the Hebrew on the next line. This is what I do on my blog. It doesn’t look very elegant, but at least I can be reasonably sure that it reads correctly:
As it happens, I’ve been working on a solution for conveying Hebrew in computerised contexts that will resolve all these issues—but more on that on another occasion…