Thanks for your question.
No, you should not register two DOIs for the same article, just because it exists first in-press and then later published in a specific issue. That’s not enough of a distinction to merit two different DOIs.
Moreover, DOIs cannot be deleted. It’s a fundamental principle of the whole DOI system that DOIs must be persistently maintained once they’re created. So, registering a ‘temporary’ DOI for the in-press version of an article would not be a viable solution. That would honestly cause far worse problems, because the workarounds we have for approximating deleting DOIs (while actually keeping them persistently maintained) are messy and complicated. You definitely don’t want to have to navigate through those processes.
The real problem here is your suffix pattern. As general guidance, we recommend that identifiable metadata like volume numbers, issue numbers, and page numbers (also journal abbreviations, author names, publication years, etc.) are not included in DOI suffixes.
DOIs are considered opaque identifiers. So, all of the meaning that’s attached to them comes from their metadata records, not from the DOI itself. We treat the DOI suffix as if it’s a totally arbitrary, random string of characters, whether or not you use a pattern with identifiable information. But, the best practice is to avoid identifiable information and use something that’s actually random and opaque.
If you change your suffix pattern going forward to something that doesn’t require knowing the eventual volume, issue, or page numbers at the point of assigning the DOI, then you can register a single DOI for the article while it’s still in in-press status. Then, once that article has later been included in an issue, you would update the metadata record of that same DOI to include the metadata that’s now available like volume, issue, and page numbers as well as a print publication date.
Please let me know if you have any further questions.