Around 10:40 pm last night, all my mongo alerts went off for one of our shards because rs.status() is now returning junk.

parse:PRIMARY> rs.status()
Wed Aug 21 16:20:05 decode failed. probably invalid utf-8 string [syncThread: 10334 Invalid BSONObj size: 247791731 (0x7300C50E) first element: �@@timezone_offset_epoch_ms: ?type=-16]
Wed Aug 21 16:20:05 why: TypeError: malformed UTF-8 character sequence at offset 78
Wed Aug 21 16:20:05 TypeError: malformed UTF-8 character sequence at offset 78 src/mongo/shell/utils.js:1018

It's returning the same junk on the primary and all secondaries.

Any idea how to fix? It's basically impossible for us to run a repairDatabase on this shard due to the 1) size of the data and 2) resetting all the padding factors would kill us.

Thanks.