Some instability which was introduced lately (no race, but bots became unstable)

https://irclog.perlgeek.de/perl6-dev/2017-05-20#i_14611238

Zoffix Looks committable6 has a race in it. If you give it code to run too fast, it hangs

Last entry below hanged until I deleted a dir in /tmp/whateverable/* with that HEAD's sha and restarted whateverable. Seems to happen when you feed it too many lines too fast:

[10:19:16] <Zoffix__> c: HEAD~20 dd BagHash.new(<a b c>).pickpairs(-200)
[10:19:18] <committable6> Zoffix__, ¦HEAD~20: «().Seq»
[10:19:28] <Zoffix__> c: HEAD~20 dd BagHash.new(<a b c>).pick (-200)
[10:19:30] <Zoffix__> c: HEAD~20 dd BagHash.new(<a b c>).pick(-200)
[10:19:31] <committable6> Zoffix__, https://gist.github.com/6087d80a0c3f6529dc2927a8e998e00e
[10:19:31] <committable6> Zoffix__, ¦HEAD~20: «("a", "b", "c").Seq»
[10:19:36] <Zoffix__> c: HEAD dd BagHash.new(<a b c>).pick(-200)
[10:19:37] <committable6> Zoffix__, ¦HEAD(e5b5d34): «().Seq»
[10:19:47] <Zoffix__> c: HEAD dd BagHash.new(<a b c>).pick(Inf)
[10:19:49] <committable6> Zoffix__, ¦HEAD(e5b5d34): «("c", "a", "b").Seq»
[10:19:51] <Zoffix__> c: HEAD dd BagHash.new(<a b c>).pick(NaN)

Start with the linked IRC log and the reported /tmp/whateverable/* directories, focusing on the bot behavior when several commands arrive rapidly. Reproduce the hang described in the transcript and determine what observable behavior would demonstrate that rapid input no longer leaves the bot unstable.