Steps to reproduce:

1. There are several triggers with nodata(24h)}=0 on different hosts, and it is already in the triggered state, as there is really no data in the data item for that time. At some point, on zabbix proxy, pollers utilization goes from 10 to 100% (this is something I have yet to figure out, but it brings up the following problem)
2. Further, it is seen that due to accumulated data, the utilization of the sender process on the zabbix proxy grows to 100%. This of course causes maximum load of trapers on the zabbix server, the server is unable to process the data stream, this can be seen by the intermittent graph from any item.
3. After this chain, triggers with nodata start to go crazy. Trigger with condition nodata(24h) or any other time, lights up and goes out every 30 seconds. If you review the history by element, there was no data at that time.

Result: