Loki does not recover from its temporary checkpoint in WAL, when Loki restarts

Hi,
When Loki restarts its WAL recovers only from its normal checkpoint and not from temporary checkpoint. Is this behavior genuine. If yes, then what is the purpose of temporary checkpoint?

It recovers first from the “checkpoint” files, which are an optimized recovery format that are created on an interval. Then, it will replay any “segments” on top of that, which are the pure-WAL entries. This idea is not too different from how Prometheus works: Prometheus TSDB (Part 2): WAL and Checkpoint | Ganesh Vernekar

Thank You owendiehl,
I agree with you that WAL recovers from checkpoint. But please look at below checkpoints and segments created in WAL:

/data/wal # ls -lrth
total 177596
drwxr-xr-x 2 root root 4.0K Feb 20 09:03 checkpoint.000000
drwxr-xr-x 2 root root 4.0K Feb 20 09:08 checkpoint.000001.tmp
-rw-r–r-- 1 root root 106.8M Feb 20 09:08 00000001
-rw-r–r-- 1 root root 66.7M Feb 20 09:11 00000002

What is the importance of checkpoint.000001.tmp? If somehow checkpoint.000000 and segments are destroyed then can WAL get recovery from checkpoint.000001.tmp? As I have checked in loki logs that recovery occurs from checkpoint.000000 but not from checkpoint.000001.tmp.