I've been trying to track down the cause of a client workflow not properly `resume`ing on a Tower installation. I've found something that reliably toggles the behavior, but what confuses me is that if this is the cause, I feel like it'd be a common complaint - yet I can't find anything. Does this ring any bells? 1. Submit workflow to Tower using

tw

CLI & providing a params YAML. Let's say one of those params is

foo = 0.0

2. Later, in Tower UI I do a

resume

on this WF &

Launch

3. In the resumed run, it's actually

foo = 0

and now the hash is different If instead in #1 I set

foo = 0

then it works AOK (which yes, means there's an obvious workaround, but that's not my question)