In case of IR issue it's a good idea to look into the media build / crazycat / dvb packages - or disable them. They also replace the rc-core subsystem which is responsible for IR remotes.
Very slow IR response with harmony remotes and current kernels <= 4.17 is mostly caused by setting command repeat to 0 (or 1). Set it to 1 (or 2) so that the remote sends at least 2 IR messages on a short button press.
The current linux media tree (which will go into 4.18) contains improvements in this area which make IR remotes a lot snappier (even Harmony remotes with command repeat disabled). I've backported these to kernel 4.14 (they are merged in LE master tree), when using kernel 4.17-rcX you can probably just pick these commits:
a86d6df84ae6 media: rc: set timeout to smallest value required by enabled protocols
284922562b81 media: rc: per-protocol repeat period and minimum keyup timer
c421c62a4a08 media: rc: mce_kbd decoder: low timeout values cause double keydowns
539327608dbe media: rc: mce_kbd protocol encodes two scancodes
63039c29f7a4 media: rc: mce_kbd decoder: fix stuck keys
cb5bd0575c41 media: rc: mce_kbd decoder: remove superfluous call to input_sync
53a62800efb2 media: rc: mce_kbd decoder: fix race condition
91352b572786 media: rc: mceusb: IR of length 0 means IR timeout, not reset
877f1a7cee3f media: rc: mceusb: allow the timeout to be configurable
aec3eadfb642 media: mceusb: MCE_CMD_SETIRTIMEOUT cause strange behaviour on device
d4589935ebab media: mceusb: filter out bogus timing irdata of duration 0
0630efeee853 media: rc: ite-cir: lower timeout and extend allowed timeout range
Display More
Read the discussion here for details: [PATCH v2 0/7] Improve latency of IR decoding — Linux media
so long,
Hias