Try to use a non-timestamp approach for CacheCollector race condition detection

Problem/Motivation

See discussion in 🐛 Race conditions/bad lock logic in CacheCollector/State Active - the write-lock logic in CacheCollector can lead to two processes locking and writing in the same millisecond, which fools CacheCollector's race condition detection because it depends on millisecond-precision timestamps.

We can potentially use a hash of the cache data instead, but attempts on that issue didn't work yet.

Steps to reproduce

Proposed resolution

Remaining tasks

User interface changes

Introduced terminology

API changes

Data model changes

Release notes snippet