System should recover if summarizer is not summarizing

[vladsud]

We observed couple files where there is an active summarize and his session is still in quorum, but it did not post any summary for 5 days / 5K ops.
While we will try to get to the bottom of it,. we should also have a recovery mechanism where another client (through quorum proposal) can take over.
This issue is about that - having recovery mechanism (and telemetry).

[arinwt]

This issue is similar to #107, but distinguishing them as follows:

#107 is for making sure we start a new summarizer when the actual summarizer client disconnects. The most common scenario is when there is only 1 client connected to a document, and somehow only the summarizer client disconnects. Currently the parent will observe this, but will not try to spawn a new summarizer client.

This issue is for handling cases when the summarizer does not necessarily disconnect, but for some reason or another the document has not had a summary for a while. It would be nice to have some mechanism for electing a different summarizer client or restarting the existing summarizer client.

[vladsud]

I'm going to close it, as we are not planning right now any work in Ignite + 30 timeframe in this space. If data suggest we need it, we will open new item.
Please feel free to push back