I feel you pain ... we have over 200 SQL Server 2000 instances that need to be monitored, and not just for SQL job failures.

The solution I went with works somewhat in the opposite direction as yours. Instead of setting up a repository server with linked servers to all the instances, I set up all the instances with a linked server to the repository server. I then configured all the instances as target servers to a master MSX server. The MSX server pushes out a SQL job to all the target servers and this SQL job collects data (including job history) and feeds it into the repository server. The repository server then processes the collected data and sends notifications as necessary. To make sure that all the target servers are actually sending their data to the repository server, they all update a time stamp as part of their collected data. I can then query the time stamps to make sure all the target servers are sending over thier data.