Type of issue

• Bug
• Enhancement
• Compliance
• Question
• Help wanted

Current Behavior

I did some benchmarks comparing v1.5.376.244 with v1.5.378.10-preview and saw more memory allocations/second (more GC pressure). Is this a known and accepted "regression"?

Before upgrade v1.5.376.244:

Client (20k variable reads per second):

Server (20k variable writes per second):

After upgrade v1.5.378.10-preview:

Client (20k variable reads per second):

Server (20k variable writes per second):

Results:

• A lot more GC memory pressure with the new version: +10 mb/s client (+50%), +20 mb/s server (+20%)
  • Might lead to higher tail latencies (e.g. higher latencies in P99 due to GC). This is maybe not visible in these benchmarks as they were run on a high performance CPU
• Overall throughput and latency is still better than previous version

(Tested on Mac Book Pro Max M4)

Expected Behavior

Lower memory consumption in variable read/write hot paths compared to previous version.

Steps To Reproduce

No response

Environment

- OS: Mac OS 26
- Environment: Mac Book Pro Max M4
- Runtime: .NET 9
- Nuget Version: v1.5.378.10-preview
- Component: 
- Server: Writes 20k variable changes per second (30k monitored items)
- Client: Reads 20k variable changes per second

Anything else?

No response