My own experience of special issues in Computer Science journals is that they are typically considered the same as regular issues. There are two types of special issues: one is (as Suresh notes) invited extended versions of the best papers from a conference. Another is just based on open calls for guest editors to propose special issues. Indexing services do not typically note that the paper was part of a special issue. I have not highlighted special issues in my CV.
Whether or not the review process is expedited depends on the guest editors. From my own sample set of journal papers, the fastest turnaround for reviews was on a special issue (three months for acceptance after one minor revision, six months to publication). However the slowest was likewise a special issue (eighteen months for acceptance after one major revision, over two years until publication).
The risk with a special issue (timewise) is that all the papers of the special issue have to be accepted so as to be published on slices of the same dead tree. So one paper that lags behind will slow down the rest of the papers from being finally published and indexed.
In terms of acceptance rates, again it can be risky. If your paper is on topic for a special issue that does not receive many other submissions, your chances of acceptance may be slightly higher since the guest editors will be anxious to fill out the special issue. Likewise if your paper was solicited (e.g., from a conference) it would stand to reason that you would have better chances since the editors have already expressed interest in publishing the paper. On the other hand, the space for a special issue is more rigidly bounded than the more elastic regular call (which spans multiple issues). Hence if you submit an unsolicited paper to a special issue that receives lots of other submissions, the special issue could be a lot more competitive than the regular issue.