diff --git a/draft/msgwork_new_imap_jobs.rst b/draft/msgwork_new_imap_jobs.rst
new file mode 100644
index 000000000..b92f29cf3
--- /dev/null
+++ b/draft/msgwork_new_imap_jobs.rst
@@ -0,0 +1,66 @@
+
+simplify/streamline mark-seen/delete/move/send-mdn job handling
+---------------------------------------------------------------
+
+Idea: Introduce a new "msgwork" sql table that looks very
+much like the jobs table but has a primary key "msgid"
+and no job id and no foreign-id anymore. This opens up
+bulk-processing by looking at the whole table and combining
+flag-setting to reduce imap-roundtrips and select-folder calls.
+
+Concretely, these IMAP jobs:
+
+    DeleteMsgOnImap
+    MarkseenMsgOnImap
+    MoveMsg
+
+Would be replaced by a few per-message columns in the new msgwork table:
+
+- needs_mark_seen: (bool) message shall be marked as seen on imap
+- needs_to_move: (bool) message should be moved to mvbox_folder
+- deletion_time: (target_time or 0) message shall be deleted at specified time
+- needs_send_mdn: (bool) MDN shall be sent
+
+The various places that currently add the (replaced) jobs
+would now add/modify the respective message record in the message-work table.
+
+Looking at a single message-work entry conceptually looks like this::
+
+    if msg.server_uid==0:
+        return RetryLater  # nothing can be done without server_uid
+
+    if msg.deletion_time > current_time:
+        imap.mark_delete(msg)  # might trigger early exit with a RetryLater/Failed
+        clear(needs_deletion)
+        clear(mark_seen)
+
+    if needs_mark_seen:
+        imap.mark_seen(msg)    # might trigger early exit with a RetryLater/Failed
+        clear(needs_mark_seen)
+
+    if needs_send_mdn:
+        schedule_smtp_send_mdn(msg)
+        clear(needs_send_mdn)
+
+    if any_flag_set():
+        retrylater
+    # remove msgwork entry from table
+
+
+Notes/Questions:
+
+- it's unclear how much we need per-message retry-time tracking/backoff
+
+- drafting bulk processing algo is useful before
+  going for the implementation, i.e. including select_folder calls etc.
+
+- maybe it's better to not have bools for the flags but
+
+  0 (no change)
+  1 (set the imap flag)
+  2 (clear the imap flag)
+
+  and design such that we can cover all imap flags.
+
+- It might not be neccessary to keep needs_send_mdn state in this table
+  if this can be decided rather when we succeed with mark_seen/mark_delete.