On the translation-i18n mailing list some translators said that they were not willing to translate the entire mailman POT file because it was so big, and many messages appeared to be in a context never seen by a mailing list subscriber.
Would it be possible for the mailman project to prioritize the messages according to the target user role?
I was told that mailman has three types of users: admins, editors, subscribers. They see different output from mailman. I would estimate that the localization of the parts that reaches subscribers is higher priority than the rest. (Because when you have 5 people or more starting a mailing list, you can expect to find one of them who understands English; this person then can become the list administrator.)
For the translators, it is acceptable to have 3 POT files instead of a single POT file; it is also acceptable to keep everything in a single POT file which has extracted tags (#. syntax) that indicate the target user (e.g. "#. to subscribers") or prioritization (e.g. "#. +", "#. ++", "#. +++").
If distinct user roles are handled by distinct source files, you can simply pass different file lists in 3 different xgettext invocations. If not, you can use different translation functions instead of _(); xgettext 0.15 will have an option to create specific extracted tags in #. syntax depending on the translation function that is used.