Usecases for deltatocumulative processor

[samin36]

We have a use case where we need to accumulate the metrics reported by multiple parallel, unique instances of a short-lived process. We cannot use the OTLP exporter to directly export to the Collector, as the most recent data point will replace the current one (instead of accumulating). Last we checked, the Collector doesn't seem to support aggregating metrics from multiple sources. To circumvent the issue, we have a custom exporter that reports the data to another long-running process, which takes care of the accumulation and ultimately exports to the Collector. Here is a diagram:

Our custom solution works, but we recently came across the deltatocumulative processor, and we are wondering if this will accomplish the same end-goal. We can adjust our short-living processes to export delta metrics instead of cumulative. The question is, does (or will) the deltatocumulative processor support aggregating metrics from multiple parallel, unique instances of a short-living process? It doesn't even have to be parallel instances. It could be the same process launched in sequence. We took a look at the current implementation, and it seems this is currently supported, but we are not sure. Supporting this use-case means the single-writer principle outline in Metrics Data Model will be violated. Any way to verify this?