Switch vs remote.send for Broadlink

Is there any difference in the effectiveness of remote.send and switches when sending Broadlink RF commands?

I have about a dozen 1st Gen Lightwave wall switches and sockets which I am controlling with RF commands from a Broadlink RM4 Pro (thus avoiding the cloud and a good deal of general crankiness).

I started off using remote.send for all of them, which worked very well. Recently I migrated all of them to switches defined in configuration.yaml, copying the command codes from the Broadlink file in .storage. This also worked very well, and made for a much less cluttered system.

Just occasionally, however, I have found that a switch or socket will become unresponsive - often enough for me to notice, not often enough to identify the problem. Not the same switch, different parts of the house, no error messages; the light just doesn’t turn on or off (although HA thinks it has - Lightwave switches don’t return their status).

I have found that if I use Developer Tools to remote.send the missed command, the light/socket responds immediately, and thereafter the switch commands start working again. (This not just a matter of HA and the lights getting out of sync, incidentally).

So what’s going on? Network issue? Broadlink thing? Imagination (quite possible)? Has anyone had similar experiences?

I have a bog standard HA setup: 2021.7.1 on a RPi 4 running Home Assistant OS.