Impact of leaving old aliases in ASM_DISKSTRING
this post deals with the bug #12860996 "QUERIES TAKE TOO LONG : V$ASM_DISK OR V$ASM_DISKGROUP" (an a couple of others on the same behaviour ,let's say a cluster startup that takes a long time as well as select against V$ASM_DISKGROUP and V$ASM_DISK that takes a long time too, waiting on the "RDBMS IPC REPLY" event).
Well, in our case, we discovered that we had "old" aliases on the asm_diskstring (such as /dev/oracle/asm*), that is links pointing to nowhere ....
After having deleted those old aliases at the OS level (rm ...), everything is OK now (I have to say that updating ASM_DISKSTRING in order to hide these old aliases is not enough, unfortunaltly).