Scarlet currently gives a log info when the SED in less than zero in any band (https://github.com/pmelchior/scarlet/blob/master/scarlet/source.py#L147-L155). While this might have been a good idea at one point, when running at scale on HSC data any blend which contains faint sources (which is most of them) has at least one source that triggers this issue, making it very difficult to parse the log files for more useful information.
Since this seems to be a common (rather than extraordinary) occurrence, I recommend only logging when all bands are negative, meaning it is a spurious detection.
Scarlet currently gives a log info when the SED in less than zero in any band (https://github.com/pmelchior/scarlet/blob/master/scarlet/source.py#L147-L155). While this might have been a good idea at one point, when running at scale on HSC data any blend which contains faint sources (which is most of them) has at least one source that triggers this issue, making it very difficult to parse the log files for more useful information.
Since this seems to be a common (rather than extraordinary) occurrence, I recommend only logging when all bands are negative, meaning it is a spurious detection.