0.1 release
  when disk is full it should log warning but not crash
  delete expired events
    * decide what to do about generalised ephemeral events
  kill plugin if it times out

0.2 release
  ? why isn't the LMDB mapping CLOEXEC
  ? plugin for stream: make sure bortloff@github didn't make a mess of it

sync
  logging of bytes up/down
  pre-calcuated tree of XOR
    full-db scan limited by since/until

features
  less verbose default logging
  multiple sync connections in one process/config
  NIP-42 AUTH
  slow-reader detection and back-pressure
  delete command
    * delete by receivedAt, IP addrs, etc
    * inverted filter: delete events that *don't* match the provided filter
  ? relay block-list events
  ? if a client disconnects, delete all its pending write messages
  ? support filtering on empty value tags

rate limits
  ! event writes per second per ip
  max connections per ip (nginx?)
  max bandwidth up/down (nginx?)
  log IP address in sendNoticeError and elsewhere where it makes sense
  ? limit on total number of events from a DBScan, not just per filter
  ? time limit on DBScan

misc
  ? periodic reaping of disconnected sockets (maybe autoping is doing this already)
  export not dying on SIGPIPE, or not getting SIGPIPE