fix: crash on iterating "delayedCalls"

[rchl]

It looks like the code allowed "loadSentry()" to run more than once and
the first run would unset "delayedCalls" causing second run to crash on
trying to iterate it.

[codecov]

Codecov Report

Merging #219 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #219   +/-   ##
=======================================
  Coverage   55.55%   55.55%           
=======================================
  Files           1        1           
  Lines          27       27           
  Branches        8        8           
=======================================
  Hits           15       15           
  Misses          9        9           
  Partials        3        3           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4fb726d...df933af. Read the comment docs.

[rchl]

Hopefully handles this issue: https://sentry.io/share/issue/2c22b5e9282a4a7986d910433e84518f/#exception

There is quite a lot of extra checking in this PR that should stop the code running twice.

But I can't quite see how this linked error was possible as for loadSentry to trigger twice (with default lazy settings), we would need to add another onNuxtReady and while there was code that did that, that function could have only been triggered by onNuxtReady hook itself so window.$nuxt should have existed already...

[rchl]

Hmm, I'm not even sure if that error is about calling forEach on undefined or about calling Sentry API from forEach. The error message is pretty poor on safari webview.

[rchl]

Will get this in and see if it still reproduces.

[rchl]

That didn't fix it...

I will have to inject some debugging code to try to figure it out.
I can only check it on production though so that makes it quite tricky to debug...

[rchl]

Just realized that this is expected. It's reposting the error caught through error handler.
The part that made me confused is the stack trace pointing at the sentry plugin. I could probably repost from a timeout to hopefully lose the stack as it's not relevant in those cases.