fix errors in log entries polling logic

[martindstone]

Log entries polling had some problems:

• interval was getting reset too often (dependency array had too many things in it)
• also it looks like the browser was caching what it thought were identical requests when we were polling and the since date had not changed -- added a random query param to bypass the cache
• multiple log entries fetches could be started at once, that should not be possible now
• logging as error when skipping polls just so that we can see it in RUM

I'd like to deploy this as a hotfix because the current behavior is pretty bad