[11.x] Avoid duplicate chain catch when nested batch fails on sync queue

[faisuc]

Prevents Bus::chain(...)->catch(...) from being invoked twice when a Bus::batch(...) inside the chain fails on the sync queue. Batch catch runs once; chain catch runs once.

As described in #55077, nested batches inside chains on the sync queue caused:

• Chain catch to run twice (once via injected batch catch, once via chain).
• Batch catch to run inside a DB transaction, rolling back its side effects on sync.

[taylorotwell]

Why does this need to be in a transaction here?

[faisuc]

Thanks, you’re right. It doesn’t need to be in a transaction on the sync path. We can’t group dispatch with DB work there anyway, and the increment is already a single atomic UPDATE. I’ve updated the PR to increment directly on sync and kept the transactional grouping for non-sync drivers.