What the difference between special issues and regular ones from scientific journal publication point of view?
Different journal may put different meaning into "special issue" It could mean that papers have been invited around a certain theme, it could mean papers come from a workshop or symposia, or that the journal have decided to gather papers around a specific topic for some reason. The point is that a special issue differs from a regular issue in that the papers constitute a collection around a theme or coming from a specific group of authors or event. Such issues can either be a decision by the journal or a request from a group of scientists to publish around "their" proposed theme. There may be additional cases but these are the ones I am familiar with within my field.
In my field (theoretical computer science), the best papers at a particular conference often get invited to a special issue of a journal. The editor of the special issue is typically one of the conference chairs, but the papers still go through the regular peer-review process of the journal, and are held to the same standards.
In addition to publishing selected papers in conferences (as mentioned by @Mangara), a publisher can announce a special issue journal to concentrate on a particular set of fields which may be a subset of its overall scope.
One main point to note is that special issues have an additional advantage that they are time-bound. That is the dates for first review result, acceptance/rejection, and publication, are all scheduled. Although sometimes, the dates may be subjected to change, this is in contrast to regular journals where the time-frame may not always be estimated.
It can vary. I have seen special issues for (1) conference, (2) theme, (3) some famous professors birthday/retirement.
In general it really doesn't matter in terms of citations, just do a full cite including issue number and the like. Nobody will notice/care about if it was special or normal. It will still get cited, abstracted, held by libraries the same way.
Sometimes the reviewing can be a little easier on special issues (they tend to have a hard time filling the issue). It's also possible that a slightly less notable paper might make it through (more "datapoint science"). Again, it's not a huge impact and shouldn't worry you, but just sharing my impression.
(Partial overlap with previous answers)