You already got a lot of good advice, but I'll chime in:

  if you're worried about confusing amanda by getting tapes out of
  sequence, don't.  flush vs dump is really more or less the same
  bookkeeping.  Many people put in the next tape and flush if there's a
  lot.

  If you are worried that you don't have enough tapes in the rotation,
  or are worried about getting more tapes, you should try really hard to
  get more tapes.  Trying to conserve tapes is rarely a good strategy.

  If some of the files are so old that you think they won't be useful,
  and you have disk space, you can just move the date-named dirs to
  another place.  I have /n1/IR-DAILY (holding dir named after my
  config), /n1/AMANDA-RESTORE (where I restore dumpfiles into), and
  /n1/AMANDA-HOLD where I put stuff I don't want flushed.  Then after
  enough time has passed that they are not useful I rm them.

  you should read up on 'autoflush'.  If your tapes are generally big
  enough and there's a little bit left, this will put it on the next
  night's dump.

  If your dumps are routinely overflowing the tape, you need to get a
  bigger drive or increase the dumpcycle.  I just went from DDS3 to LTO2
  and wish I had done it earlier.  The $5K has already paid for itself
  in saved labor and better backups.

  If you flush a lot, then you should think about how your dumpcycle and
  #tapes relate.  I try to set dumpcycle to ensure 2 fulls of each DLE
  in the rotation.  If you flush onto 2 tapes every day, then you may
  need to set dumpcycle shorter because tapes are used for flushing that
  amanda isn't expecting to dump on.  If you're just behind now but not
  falling behind routinely that's not such a big deal, esp. with 50
  tapes.

  run 'amadmin balance' and compare balanced size to your tape.  If it's
  more than 50%, or maybe even > 25%,you need a bigger drive or a bigger
  dumpcycle/tapecycle.  At 50 tapes, the only reasonable answer is a
  bigger drive, assuming dumpcycle is in the 15-20 range which is where
  I'd put it for 50 tapes.