Dev mode stops without reason instead of restart when fired by code Changed (worked with 1.4.2.Final)

[JiriOndrusek]

Describe the bug
Problem can be simulated on the example from camel-quarkus: https://github.com/apache/camel-quarkus/tree/master/examples/timer-log

Problem doesn't happen with quarkus 1.4.2.Final. (but happens with 1.5.0.cr1, 1.5.0.Final and master)

Expected behavior
App is reloaded with custom code change.

Actual behavior
App stops.

To Reproduce
Steps to reproduce the behavior:

1. Download https://github.com/apache/camel-quarkus/tree/master/examples/timer-log (version of quarkus could be chaged in https://github.com/apache/camel-quarkus/blob/master/pom.xml#L69)
2. Run mvn compile quarkus:dev
3. Change message in TimeRoute.java

Additional context

Problem seems to be happening somewhere near https://github.com/quarkusio/quarkus/blob/master/core/runtime/src/main/java/io/quarkus/runtime/util/ClassPathUtils.java#L84
I can debug that process enters this method with parameter META-INF/quarkus-config-roots.list but it never leaves (no exception is thrown), Java ends also.
I've tried to debug issue also in https://github.com/quarkusio/quarkus/blob/master/independent-projects/bootstrap/core/src/main/java/io/quarkus/bootstrap/classloading/QuarkusClassLoader.java#L154,
but it usually ens in different places. I'm probably missing something here.

[JiriOndrusek]

More facts:

• I've tried Runtime.addShutdownHook(Thread thehook) and it is executed in 1.5.x not in 1.4.2
• I've tried @observes StartupEvent ev and @observes ShutdownEvent ev - in 1.5.x only shutdown event is observed (in 1.4.2 both)

Important fact (from my POW) is that message from Runtime.addShutdownHook is printed before ClassPathUtils (from the description) are executed. It seems that problem is probably not in classLoading as I suggested in description. App is stopped but not started again.

[JiriOndrusek]

Problem is caused by following change:
57ca8e8

[JiriOndrusek]

Simple solution is to add LockSupport.park(); (see 57ca8e8#diff-10e19a8856031bf0bc6780fd924c1091L58)
I don't see possible impact and don't want to revert original change/fix. May I ask you your opinion about this, @stuartwdouglas ?