actually uphold the bind_host parameter. in theory should make ipv6 binds work like bind_host: 'fe80::1%lo0'

Metrics

Option to serve metrics from their own localhost-only TCP port instead of muxed on the main listener

Bugfix to sql_txn_timer increment - add only the per-TXN duration, not the total time ever spent since boot

Replace the @metrics.counted annotations in federation with specifically-written counters and distributions

Rename the timer metrics exported by synapse.storage to append _time, so the meaning of ':total' is clearer

Export CacheMetric as hits+total, rather than hits+misses, as it's easier to derive hit ratio from that

Have all @metrics.counted use a single metric name vectored on the method name, rather than a brand new scalar counter per counted method

Add TimerMetrics to shadow the PerformanceCounters in synapse.storage; with the view to eventually replacing them entirely

Put some gauge metrics on the number of notifier listeners, and notified-on objects (users, rooms, appservices)

Neater implementation of metric render methods by pulling out 'render' as a base method that calls self.render_item