sweeptimelockmanual fails with "target script not derived" & flags don't appear to widen search range

[jason-me]

I'm trying to sweep a locally-force-closed channel's timelocked output using sweeptimelockmanual. This exact method worked successfully on a sibling channel from the same node moments earlier, but fails on this one.

Command run:

chantools sweeptimelockmanual \
  --frombackup ~/current.backup \
  --channelpoint 599b091845fd65a0e15c9bd704847bc0253b8f50c5b761e466d2de24e0d0c349:0 \
  --timelockaddr bc1qavc4vv7ncn5sdf7337m26duzwkj6jfr9jhr875gp3lsmpklarxxsh9k9md \
  --sweepaddr <redacted> \
  --feerate 5 \
  --maxnumchanstotal 1000 \
  --maxnumchanupdates 5000 \
  --maxcsvlimit 2016 \
  --publish

Output:

Starting to brute force the time lock script, using: remote_rev_base_point=0222c416eda336a2ce5178c9406bc64a77d6e4ad9207c3672a4e91e2134142a0d5, start_csv_limit=144, max_csv_limit=145, start_num_channels=331, max_num_channels=332, max_num_chan_updates=5000
Tried 332 of 500 keys.
Error: target script not derived

Notable issue: Despite passing --maxnumchanstotal 1000, --maxcsvlimit 2016, the debug output still shows the same narrow range as the default run (start_csv_limit=144, max_csv_limit=145, start_num_channels=331, max_num_channels=332) and these flags don't appear to be taking effect.

Context:

chantools version: v0.14.1
The channel was originally force-closed by lnd itself (ChanStatusLocalCloseInitiator), commitment type ANCHORS, csv_delay reported as 144 by lncli
Channel point: 599b091845fd65a0e15c9bd704847bc0253b8f50c5b761e466d2de24e0d0c349:0
The remote revocation base point was correctly identified from the channel.backup file
Timelock output confirmed on-chain 8+ months ago, well past any CSV maturity

[guggero]

When you use --frombackup, then the CSV and channel start/end numbers don't need to be guessed, as they're in the SCB file.
Which speeds up the search massively.

So I assume the --maxnumchanupdates 5000 is not sufficient. If it's an old channel or a lot of payments (or rebalances) have been made, then that number can reach a million or more in some cases. So that's the number you'll want to tune.
Because all other parameters are fixed, it shouldn't take super long to test very high values.

[jason-me]

Thanks! I will try it again soon. Very helpful.