OTEL_EXPORTER_OTLP_TIMEOUT is in seconds instead of milliseconds, not following semantic conventions

It seems that the OTEL_EXPORTER_OTLP_TIMEOUT environment variable in the OpenTelemetry Ruby SDK is interpreted in seconds, rather than milliseconds as specified in the official semantic conventions.

This inconsistency can create problems in polyglot environments, where the same environment variable is injected across services written in different languages.

For example, setting OTEL_EXPORTER_OTLP_TIMEOUT=10000 (expecting 10 seconds) would result in a very large timeout in Ruby (10,000 seconds) but the correct timeout in other languages.

I understand that changing this behavior would be a breaking change, but I wanted to check if there are any plans to bring this in line with the semantic convention in the near future. Even an opt-in behavior or deprecation warning could help reduce confusion.

Let me know your thoughts — just trying to align things on our end and avoid unexpected issues.

Thanks in advance !

Start by tracing how the OpenTelemetry Ruby SDK handles OTEL_EXPORTER_OTLP_TIMEOUT and compare its unit with the linked semantic convention. Determine the appropriate compatibility or migration behavior, then verify that a value such as 10000 is handled consistently with the documented milliseconds convention.