With the new method of splitting out the log messages for all different levels, it makes it particularly difficult to correlate past log events since the logs rollover at different times.
Suggestion is to compress the logs to keep much more data under the 100mb limit.
Would it make sense to only have 2 log files for each type? 1 "in use" log file that's not compressed, and then roll it over into a single aggregate historical log file that is compressed? That would make it a little easier to look at the log files while on the node itself rather than using the collect_info to aggregate them together...