MatthewFlamm

In my testing, I'm often getting no data even in the first request and random no data returned for various time periods. I don't see much rhyme or reason other...

> The no data problem seems to have gone away if I change the headers > [MatthewFlamm/pynws#201](https://github.com/MatthewFlamm/pynws/pull/201) This will likely be a temporary fix until the server starts messing up...

One thought for feedback... We keep track of the last successful update times (including whether data is returned) in the `entry.runtime_data`. Then we only raise `UpdateFailed` in the coordinator call...

Most significant changes in https://github.com/home-assistant/core/pull/117109/commits/8b99d0e98464bd9fbd01399748b98c398749e160 include using a custom `DataUpdateCoordinator` for the observations. This enables us to internally raise an `UpdateFailed` only after a certain amount of time elapses from...

> If there are no breaking changes in the library bump, please do that in a separate PR first. Thanks! https://github.com/home-assistant/core/pull/117820

One unresolved pain point in this implementation is that if the first update fails, it will wait a full 10 minutes to make another update request. To keep the PR...

> Looks like its better but still seems to be as long as 90 minutes before new data is returned (maybe longer) We can't fix the server not returning valid...

A problem with the current use of `call_with_retry` that wraps any `update_*` call is that is is harder to make special cases for `update_observation` for example.

One thought that I like is to have a new keyword arg for each `update_*` method that would raise an error when some condition is met for the data coming...

Observations already allow a `start_time` parameter, right now we trust NWS to satisfy this request. We should also filter it if user supplies this and NWS returns old data.