Some time ago I built an MFJ Cub 20m transceiver. It was fun to build, and has a very nice receiver with true single sideband selectivity using a crystal filter. I added a built-in rechargeable battery pack using ten Toshiba AAAL NiMH cells I got from a RadCom advertiser, making a very small self-contained transceiver, just add an earpiece, key and antenna. Unfortunately the VFO drifted so much I was too embarrassed to use it on the air, so it gathered dust on a shelf for several years.
The other day I decided to try to cure the drift by replacing the two polystyrene capacitors supplied by MFJ, along the lines suggested in an article by Larry East, W1HUE. First I tried using a pair of ordinary disc ceramic capacitors, but these made the drift far worse, though in the opposite direction. So I ordered some NP0 miniature ceramic capacitors. These show no change in value when warmed or cooled according to my capacitance meter. I installed them, and the drift is appreciably reduced, though the transceiver still does drift slowly LF when first switched on. I haven't left it on to see if it eventually stabilizes. But I suspect that whatever causes the drift it is not those two capacitors.
I was realigning the transceiver with the antenna connected, and heard Evgeny, UA1TET from near Novogorod calling CQ. I replied and he came back to my first call. He gave my signal 539 and complimented me on my signal with 1 watt. OK, European Russia is not exactly DX, but to make the contact with 1 watt to a dipole zigzagged in the attic is still a thrill. QRP is just amazing!