George Gensure

That change to the visitor for queue lifetime looks pretty good - can comment further if you put up a PR for it. Wasn't aware of that leak, and I've...

So the value to be compared against the limit is "the sum of all file sizes for the tree of inputs to an action"?

"the digest" as a subject for comparison does not narrow the scope of the filter - we already limit by size generally, and ensure that no blobs can exist in...

The premise of this change doesn't line up with the flow control - you can't write any bytes to the file without exclusivity. You cannot request new bytes from the...

There's a hitch here: The client isn't saying anything about the actual ttl value to the remote cache. We can probably assume that a default coupled pair of client/remote cache...

> For the redis issue, today I can post up the stack that I was hitting. I've been doing an incremental rollout and scaling it in real time as we...

That NOT_FOUND was supposed to inspire the client to retry the execution (in the virtuous loop in RemoteSpawnStrategy, outside of executeRemotely). This inspiration has fallen flat, as bazel obviously sees...

Can you provide some context on the reason for this? What is the benefit to applying a non blocking uuid gen?

> Can you provide some context on the reason for this? What is the benefit to applying a non blocking uuid gen? Do you really have enough write generation that...

> Good job profiling this. It's nice to see you got a performance boost as well. My only suggestion is to make the UUID fetching something like a [Supplier](https://docs.oracle.com/javase/8/docs/api/java/util/function/Supplier.html). That...