Sergey Vasilyev

Hello, @paxbit. Yes, you might be right with this behaviour. Kopf has some protection measures against quickly arriving events — the "batch window" (`settings.batching.batch_window`, defaults to 0.1 seconds). All events...

> Errm, yes I'm quite aware of that and I actually thought about mentioning this but thought it'd be too obvious ;) Sorry, I didn't have any intention to question...

1. Yes, `batch_window=None` seems like a good option for those who want it. The docstring must explain the consequences briefly. 2. I didn't fully get what is happening in this...

As a side-note: the latency can be added by using [toxiproxy by Shopify](https://github.com/Shopify/toxiproxy). ``` toxiproxy-server toxiproxy-cli create kopf_k3s -l localhost:11223 -u localhost:50016 toxiproxy-cli toxic add kopf_k3s -t latency -a latency=3000...

In response to the comments: To no.1. I agree in principle. Though, I have some doubts about the implementation. Per-resource configuration looks questionable to me. The batching window is a...

To no.2: A similar approach was once offered for some other unrelated problem. There are reasons why patches should not be handled any other way than by applying them to...

_Notes for myself:_ --- It seems that ToxiProxy cannot help here. It slows down all requests, including patching. If patching is slowed down, the operator blocks for the whole duration...

@paxbit Thanks for reporting this issue with the details. That let me fully understand the issue directly from its description, without a repro. Indeed, this might be the case. May...

Good news! In #784, a similar issue was reported with event loss. Unlike here, it was happening with no big scale, at a relatively small scale, but with multiple synchronous...

Hello. Thanks for reporting. I tried to reproduce the issue with your code snippet, and I could not. Everything works as expected: the deleted resource disappear from the index. Can...