Describe your environment Windows 10, VS 2019, Bazel 5.0.0

Steps to reproduce

1. Build: bazel build :prometheus_example from examples/prometheus
2. Run the example: ..\..\bazel-bin\examples\prometheus\prometheus_example.exe histogram
3. Get the metrics: curl localhost:8080/metrics

What is the expected behavior?
Output consistent with Prometheus histogram metrics specification:

1. one entry for prometheus_metric_example_sum
2. one entry for prometheus_metric_example_count
3. one entry for each value of label "le" of prometheus_metric_example_bucket (0, 5, 10, etc. up to "+Inf")

What is the actual behavior?
Many entries of each metric, often not with the same value even though the timestamp is the same. E.g.:

prometheus_metric_example_sum 2520.7 1654623992100
prometheus_metric_example_sum 1058.2 1654623992100
prometheus_metric_example_sum 0 1654623992100

This is just an extract, there were exactly 50 entries when I counted them by piping to wc -l. The Prometheus exposition format specification states "Each line must have a unique combination of a metric name and labels. Otherwise, the ingestion behavior is undefined."