Fluent Bit vs OpenSearch Dashboards

Development runs on parallel trains: 5.0.x with 5.1.0 just tagged, and 4.2.x as the maintenance line that receives backports of the same fixes weeks later. The content is overwhelmingly defensive — out-of-bounds reads in the snappy framed parser, a 100% CPU hang and use-after-free in in_forward, memory leaks in out_stackdriver, endian-safe gzip magic-number reads, fuzzer hardening. New capability shows up in small increments: user-defined Windows performance counters via PDH, eBPF exec tracing, IANA timezone support for native timestamps, OTLP resource grouping preserved through the pipeline.

This is a project spending its effort on the reliability of a component that sits in the critical path of everyone's logging. The recurring pattern — parser and protocol handling fixes, memory safety in output plugins, fuzz-test infrastructure — reads as a sustained hardening campaign rather than feature development, with OpenTelemetry and Windows the two places where the surface is still growing.

Expect 5.1.x to build out while 4.2.x continues receiving the same fixes as backports, and for OTLP handling and Windows metrics inputs to remain where new options appear. The 5.1.0 note carries only a test commit, so what distinguishes the new minor line is not yet visible from the changelog.