01
Know what the numbers represent
The writer maps each 8-bit channel value into a value between zero and one and stores it as a half-float. It does not perform a scene-linear color transform or carry an alpha channel. A compositor that assumes scene-linear EXR values may therefore display the picture differently unless you assign the appropriate interpretation in that application.
02
Use a diagnostic ramp
Make a small source with black, mid-gray, and white patches. A channel value of 128 is written at approximately 128 divided by 255, subject to half-float rounding. Inspect those patches numerically in your receiving application and compare its viewer transform. This is more informative than judging an EXR solely by the thumbnail a file browser happens to show.
03
Keep expectations tied to the source
This route suits format compatibility experiments and reference pictures. It is not a substitute for exporting a render, camera RAW development, depth pass, or HDR image from the software that created it. The uncompressed output can be considerably larger than a JPG. Use a modest image for your first import and keep the original color-managed source.