Otherwise it will be a huge mess for people to re-remember task numbers.
If this issue is not fixed:
* X' person-hours of work would be wasted by N frequent bugzilla users (including developers) to unlearn and re-learn issue numbers, for a cost of X dollars;
* Y dollars of costs are taken in terms of risks for that links and references to bugs break, because we're not able to redirect everything or rewrite all documents produced or printed in the last 10 years worldwide (not just inbound links, it's also in the content of MediaWiki:CodeReview, Gerrit commit messages and discussions, change logs, engineering reports and of course within the content of the bugs themselves);
Fixing this issue may cost W dollars. W is arbitrarily assumed to be lower than the sum above.
Preserving Bugzilla report numbers in Phabricator is a very complex task. People used to memorize bug numbers can find the new tasks by using the old Bugzilla URLs (probably in their browser history already) or Maniphest's advanced search (see T991).
Why the complexity:
* We have close to 1000 tasks created in phabricator.wikimedia.org. In order to make Bug 1 T1 we would need to change the current task numbers or hack Phabricator with a local patch in order to create new series of numbers i.e. "B1".
* In order to migrate 70.000 bugs in a decent time-frame (1-3 days), we need to do migrate bugs simultaneously in batches. In order to keep bug numbers, we would need to do the migration strictly sequential, starting to migrate a new bug only after the previous one has been migrated. This would increase dramatically the time required for the migration, with Phabricator and Bugzilla down.
* Guaranteeing that all tasks would preserve the bug number is very risky. In the way Phabricator creates tasks via API, one error producing an incomplete attempt would create a new task that would take the next number available, breaking the correlation of bug numbers from that point.
While we agree that the change of numbers will be annoying for some people with good memory, we believe that --with a bit of patience-- the problem will be solved as new numbers become as relevant as the old ones were. In the meantime, the transition should be simpler with the automatic redirects from Bugzilla URLs to Phabricator tasks, and probably even simpler when combined with the users' browser history.