This is for a grab-bag of minor issues, none of which really deserve a ticket on their own.

• lang/python/wiredtiger.i has a comment about needing an explicit conversion for plh_put. We should either figure out why we need the conversion, or edit the comment to sound less like a question.

• The comment for wt_conn_remove_page_log should talk about removing all added page_log entries, rather than a single page_log.

• We need to clarify the thread safety situation. The documentation (in wiredtiger.in says PALI implementations are responsible for protecting all shared data from races. But palm->reference_count isn't atomic, and palm_terminate seems to rely on it being atomic – e.g. where it says "We should be single threaded at this point". We should also specify if the checkpoint interfaces need to protect shared data.

• palm_extension_init has a forward declaration immediately before the function body, which is redundant.

• Some of the logic for palm_get_last_lsn is weird – we only ever store values 1 or greater for PALM_KV_GLOBAL_REVISION, so the check should probably be an assert instead.

• palm_get_complete_checkpoint should convert MDB_NOTFOUND to WT_NOTFOUND where it calls palm_kv_get_global, just like palm_get_complete_checkpoint_ext does. There are other cases where we miss this, so we should check those cases too.

• We should check if we can combine the various methods we use to get the last checkpoint – maybe palm_get_complete_checkpoint can be a wrapper for palm_get_complete_checkpoint_ext.