Well we go our issue taken care of. It was not pretty. Here is the lowdown. Essentially we overloaded replication with an update that affected 7 million rows, along with several other process that were updating lots of rows. So we stopped replicating some publications which basically left all the updated rows in the transaction log as marked for replication. The transaction marked for replication could not be removed from the log during normal backup. This created problems because the logs were growing taking up disk space.

Normally in this situation executing sp_repldone will clean these up. Problem was we had so many transactions that sp_repldone was running for hours on small databases.

This left us with one option. Detach the database, rename the log file, reattach the database. A final approach but it worked and the server is functioning better than before.

Hopefully this can help someone else.

Ross