We have had several instances where building large indexes causes the primary to become inaccessible and step-down to secondary. This has happened during a copyDatabase command and when it stepped down the copy halted and the primary went into rollback then secondary.

This happened several times during the copyDatabase and also just during an ensureIndex. I backed one environment we were getting this error on back to 2.0.1 and it completed normally. That part is anecdotal but it did work in that one case. Also, I don't recall seeing this before our upgrade to 2.0.2.