Comment 3 for bug 1454891

In the initial paste (private link): https://pastebin.canonical.com/131827/
log line attempted (677k) over max size(10k), printing beginning and end ... update juju.txns.stash query: { _id: { c: "lease", id: "trusty-server-leadership" }, n: "d5b7100b" } update: { $set: { txn-queue: [ "55352def91a7b10f460007b4_127104b4", "55352e1091a7b10f46000802_01edfd48",...

That looks like something that is having severe problems with conflicting transactions on the trusty-server-leadership document. Enough so that it ends up with hundreds/thousands? of txn-queue ids (these are transactions that want to be applied to the document, but are unable to be cleaned up for some reason.)

Also on the document:
txn-revno: -37441
I'm pretty sure txn-revno is supposed to stay a positive value. And negative values are used to toggle some sort of boolean state.
https://github.com/go-mgo/mgo/blob/v2/txn/flusher.go#L481

Remove operations try to invert the revno, as do Insert operations (Insert because it should start at -1), and Update operations are supposed to just increment it.

So maybe one of the txns is trying to Remove the document, and a bunch of others are trying to Update it?