A single update command may contains several update op entries. We should collect this number as one of the metrics that we collect for update command query stats.

db.runCommand({
    update: coll,
    updates: [
        {q: {...}, u: {...}, ...}, // one of many update op entries
        {q: {...}, u: {...}, ...},
        {q: {...}, u: {...}, ...},
        {q: {...}, u: {...}, ...},
        ...
    ],
    ...
})

 
Our approach in this project has been to collect statistics at the update op entry level, not for the update command as a whole. But it seems useful to know the number of update op entries in the command as a whole, and store that number in the stats for each update op entry.

This strategy should give us an idea of how many entries are in each update command. For commands that have very large numbers of entries, our numbers might be skewed if we record stats for more than one entry in the command. But this seems like a very rare case.