streamly | A streaming REST client for Ruby , using libcurl

In the code above, parallely decided to create one Haskell thread for each element in the list getAggregateIds, which is [1,2,3,1,2,3]. parallely does not care about there being some duplicate elements in the list: it simply starts a thread for each one.

In principle, parallely could allocate only a small number of Haskell threads and reuse them later on (possibly for the same duplicate ID, or another one), but there would be no performance gain in doing so. Indeed, the crucial part here is that an Haskell thread is being allocated, not an OS thread,

Haskell threads are very lightweight, they use very little memory and so they are very cheap to create and dispose. Trying to reuse them would possibly lead to worse performance.

Further, the Haskell runtime can execute many Haskell threads in a single OS threads. Usually, a small pool of OS threads is kept around by the runtime, and Haskell threads are mapped to those. Since OS threads are not as lightweight OS threads are indeed reused between Haskell threads.

Finally, note that the ThreadId is the name of the Haskell thread, not the OS one, so it's normal to see no reuse of those IDs.

you container is resizing, but the img is overflowing.

You need to add width:100% to your img, so it resizes to whatever width the container has.