More control over when buffer gets flushed

[msaspence]

Currently the buffer is flushed every 5 seconds, more complex models may be nice to have:

• max buffer time - if buffer hasn't been flushed by this it will be
• min buffer time - after a metric is tracked the script would wait this amount of time before flushing buffer, if a new metric is tracked in that time it will delay it another min buffer time
• min buffer size - the buffer won't flush it is smaller than this
• max buffer size - the buffer will flush at this size regardless of other considerations

[theninj4]

This sounds great - we could set sensible defaults and let users override as necessary. These changes may need to come after some of the other network optimisation stuff, right now the numbers are aligned to mitigate any possible network performance issues. It could be possible to tweak the above options and introduce performance issues.

[msaspence]

This functionality could also be loaded in a second phase, i.e. the embedded script has a super simple implementation to save bytes. A "phase two" script could replace this with an implementation with more options.